fake-academic-paper-generation/transformer-training-logs.txt

WARNING: Logging before flag parsing goes to stderr.
W0804 19:23:44.963335 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/expert_utils.py:68: The name tf.variable_scope is deprecated. Please use tf.compat.v1.variable_scope instead.

W0804 19:23:47.298877 140200711067520 lazy_loader.py:50] 
The TensorFlow contrib module will not be included in TensorFlow 2.0.
For more information, please see:
  * https://github.com/tensorflow/community/blob/master/rfcs/20180907-contrib-sunset.md
  * https://github.com/tensorflow/addons
  * https://github.com/tensorflow/io (for I/O related ops)
If you depend on functionality not listed there, please file an issue.

W0804 19:23:48.913978 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/adafactor.py:27: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.

W0804 19:23:48.914525 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/multistep_optimizer.py:32: The name tf.train.AdamOptimizer is deprecated. Please use tf.compat.v1.train.AdamOptimizer instead.

W0804 19:23:48.933969 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/mesh_tensorflow/ops.py:4237: The name tf.train.CheckpointSaverListener is deprecated. Please use tf.estimator.CheckpointSaverListener instead.

W0804 19:23:48.934196 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/mesh_tensorflow/ops.py:4260: The name tf.train.SessionRunHook is deprecated. Please use tf.estimator.SessionRunHook instead.

W0804 19:23:48.985285 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/rl/gym_utils.py:219: The name tf.logging.info is deprecated. Please use tf.compat.v1.logging.info instead.

W0804 19:23:49.096743 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/trainer_lib.py:109: The name tf.OptimizerOptions is deprecated. Please use tf.compat.v1.OptimizerOptions instead.

W0804 19:23:49.880575 140200711067520 deprecation_wrapper.py:119] From /usr/local/bin/t2t-trainer:32: The name tf.logging.set_verbosity is deprecated. Please use tf.compat.v1.logging.set_verbosity instead.

W0804 19:23:49.880771 140200711067520 deprecation_wrapper.py:119] From /usr/local/bin/t2t-trainer:32: The name tf.logging.INFO is deprecated. Please use tf.compat.v1.logging.INFO instead.

W0804 19:23:49.880886 140200711067520 deprecation_wrapper.py:119] From /usr/local/bin/t2t-trainer:33: The name tf.app.run is deprecated. Please use tf.compat.v1.app.run instead.

I0804 19:23:49.881355 140200711067520 usr_dir.py:43] Importing user module t2t_paper_generation_problem from path /content/fake-academic-paper-generation
W0804 19:23:49.883288 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/hparams_lib.py:49: The name tf.gfile.Exists is deprecated. Please use tf.io.gfile.exists instead.

W0804 19:23:49.883560 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/trainer_lib.py:780: The name tf.set_random_seed is deprecated. Please use tf.compat.v1.set_random_seed instead.

W0804 19:23:49.890107 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/bin/t2t_trainer.py:282: The name tf.gfile.MakeDirs is deprecated. Please use tf.io.gfile.makedirs instead.

I0804 19:23:49.890470 140200711067520 t2t_trainer.py:286] Generating data for paper_generation_problem
W0804 19:23:49.892291 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/data_generators/generator_utils.py:164: The name tf.python_io.TFRecordWriter is deprecated. Please use tf.io.TFRecordWriter instead.

I0804 19:23:49.905993 140200711067520 generator_utils.py:232] Downloading https://github.com/lipanpanpanpan/fake-academic-paper-generation/raw/master/dataset/preprocessed_data.txt to experiment/transformer/transformer_small/tmp/paper_dataset.txt
W0804 19:23:49.906136 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/data_generators/generator_utils.py:234: The name tf.gfile.Copy is deprecated. Please use tf.io.gfile.copy instead.

100% completed
W0804 19:23:51.389821 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/data_generators/generator_utils.py:242: The name tf.gfile.Rename is deprecated. Please use tf.io.gfile.rename instead.

I0804 19:23:51.390216 140200711067520 generator_utils.py:247] Successfully downloaded paper_dataset.txt, 38382903 bytes.
I0804 19:23:51.463192 140200711067520 generator_utils.py:170] Generating case 0.
I0804 19:23:59.097003 140200711067520 generator_utils.py:170] Generating case 100000.
I0804 19:24:06.568216 140200711067520 generator_utils.py:170] Generating case 200000.
I0804 19:24:14.001129 140200711067520 generator_utils.py:193] Generated 299861 Examples
I0804 19:24:14.003109 140200711067520 generator_utils.py:527] Shuffling data...
W0804 19:24:14.003282 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/data_generators/generator_utils.py:469: tf_record_iterator (from tensorflow.python.lib.io.tf_record) is deprecated and will be removed in a future version.
Instructions for updating:
Use eager execution and: 
`tf.data.TFRecordDataset(path)`
W0804 19:24:14.027454 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/data_generators/generator_utils.py:513: The name tf.gfile.Remove is deprecated. Please use tf.io.gfile.remove instead.

I0804 19:24:16.187735 140200711067520 generator_utils.py:530] Data shuffled.
W0804 19:24:16.190629 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/trainer_lib.py:121: The name tf.GraphOptions is deprecated. Please use tf.compat.v1.GraphOptions instead.

W0804 19:24:16.190868 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/trainer_lib.py:127: The name tf.GPUOptions is deprecated. Please use tf.compat.v1.GPUOptions instead.

W0804 19:24:16.191073 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/trainer_lib.py:240: RunConfig.__init__ (from tensorflow.contrib.learn.python.learn.estimators.run_config) is deprecated and will be removed in a future version.
Instructions for updating:
When switching to tf.estimator.Estimator, use tf.estimator.RunConfig instead.
I0804 19:24:16.191280 140200711067520 trainer_lib.py:263] Configuring DataParallelism to replicate the model.
I0804 19:24:16.191368 140200711067520 devices.py:76] schedule=continuous_train_and_eval
I0804 19:24:16.191456 140200711067520 devices.py:77] worker_gpu=1
I0804 19:24:16.191523 140200711067520 devices.py:78] sync=False
W0804 19:24:16.191632 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/devices.py:139: The name tf.logging.warn is deprecated. Please use tf.compat.v1.logging.warn instead.

W0804 19:24:16.191702 140200711067520 devices.py:141] Schedule=continuous_train_and_eval. Assuming that training is running on a single machine.
I0804 19:24:16.192488 140200711067520 devices.py:170] datashard_devices: ['gpu:0']
I0804 19:24:16.192576 140200711067520 devices.py:171] caching_devices: None
I0804 19:24:16.193089 140200711067520 devices.py:172] ps_devices: ['gpu:0']
I0804 19:24:16.193797 140200711067520 estimator.py:209] Using config: {'_task_type': None, '_task_id': 0, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7f82b6e3b470>, '_master': '', '_num_ps_replicas': 0, '_num_worker_replicas': 0, '_environment': 'local', '_is_chief': True, '_evaluation_master': '', '_train_distribute': None, '_eval_distribute': None, '_experimental_max_worker_delay_secs': None, '_device_fn': None, '_tf_config': gpu_options {
  per_process_gpu_memory_fraction: 1.0
}
, '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_secs': None, '_log_step_count_steps': 100, '_protocol': None, '_session_config': gpu_options {
  per_process_gpu_memory_fraction: 0.95
}
allow_soft_placement: true
graph_options {
  optimizer_options {
    global_jit_level: OFF
  }
}
isolate_session_state: true
, '_save_checkpoints_steps': 1000, '_keep_checkpoint_max': 20, '_keep_checkpoint_every_n_hours': 10000, '_model_dir': 'experiment/transformer/transformer_small/output', 'use_tpu': False, 't2t_device_info': {'num_async_replicas': 1}, 'data_parallelism': <tensor2tensor.utils.expert_utils.Parallelism object at 0x7f82b6e3b4e0>}
W0804 19:24:16.194043 140200711067520 model_fn.py:630] Estimator's model_fn (<function T2TModel.make_estimator_model_fn.<locals>.wrapping_model_fn at 0x7f82b6e3f7b8>) includes params argument, but params are not passed to Estimator.
W0804 19:24:16.194649 140200711067520 trainer_lib.py:724] ValidationMonitor only works with --schedule=train_and_evaluate
I0804 19:24:16.197000 140200711067520 estimator_training.py:186] Not using Distribute Coordinator.
I0804 19:24:16.197229 140200711067520 training.py:612] Running training and evaluation locally (non-distributed).
I0804 19:24:16.197569 140200711067520 training.py:700] Start train and evaluate loop. The evaluate will happen after every checkpoint. Checkpoint frequency is determined based on RunConfig arguments: save_checkpoints_steps 1000 or save_checkpoints_secs None.
W0804 19:24:16.206409 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/training/training_util.py:236: Variable.initialized_value (from tensorflow.python.ops.variables) is deprecated and will be removed in a future version.
Instructions for updating:
Use Variable.read_value. Variables in 2.X are initialized automatically both in eager and graph (inside tf.defun) contexts.
I0804 19:24:16.217053 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-train*
I0804 19:24:16.223310 140200711067520 problem.py:670] partition: 0 num_data_files: 100
W0804 19:24:16.225496 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/data_generators/problem.py:680: parallel_interleave (from tensorflow.python.data.experimental.ops.interleave_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.data.Dataset.interleave(map_func, cycle_length, block_length, num_parallel_calls=tf.data.experimental.AUTOTUNE)` instead. If sloppy execution is desired, use `tf.data.Options.experimental_determinstic`.
W0804 19:24:16.431521 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/data_reader.py:37: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.cast` instead.
W0804 19:24:16.471996 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/data/experimental/ops/grouping.py:193: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.where in 2.0, which has the same broadcast rule as np.where
W0804 19:24:16.536600 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/data_reader.py:231: The name tf.summary.scalar is deprecated. Please use tf.compat.v1.summary.scalar instead.

W0804 19:24:16.547811 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/data_reader.py:233: to_float (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.cast` instead.
I0804 19:24:16.582735 140200711067520 estimator.py:1145] Calling model_fn.
I0804 19:24:16.595087 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'train'
W0804 19:24:16.675199 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/t2t_model.py:243: The name tf.summary.text is deprecated. Please use tf.compat.v1.summary.text instead.

I0804 19:24:17.482010 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 19:24:17.951250 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 19:24:18.102306 140200711067520 t2t_model.py:2172] Building model body
W0804 19:24:18.335082 140200711067520 deprecation.py:506] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/models/transformer.py:156: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.
Instructions for updating:
Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.
W0804 19:24:18.372541 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/layers/common_layers.py:3106: The name tf.layers.Dense is deprecated. Please use tf.compat.v1.layers.Dense instead.

W0804 19:24:18.797479 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/layers/common_attention.py:1217: The name tf.summary.image is deprecated. Please use tf.compat.v1.summary.image instead.

I0804 19:24:19.331498 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
W0804 19:24:19.461583 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/learning_rate.py:107: The name tf.train.get_or_create_global_step is deprecated. Please use tf.compat.v1.train.get_or_create_global_step instead.

I0804 19:24:19.462944 140200711067520 learning_rate.py:29] Base learning rate: 2.000000
I0804 19:24:19.473967 140200711067520 optimize.py:327] Trainable Variables Total size: 1644032
I0804 19:24:19.474202 140200711067520 optimize.py:327] Non-trainable variables Total size: 5
I0804 19:24:19.474358 140200711067520 optimize.py:182] Using optimizer adam
I0804 19:24:21.590090 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 19:24:21.591542 140200711067520 basic_session_run_hooks.py:541] Create CheckpointSaverHook.
I0804 19:24:22.326212 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 19:24:22.340769: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 2300000000 Hz
2019-08-04 19:24:22.342739: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x9512540 executing computations on platform Host. Devices:
2019-08-04 19:24:22.342773: I tensorflow/compiler/xla/service/service.cc:175]   StreamExecutor device (0): <undefined>, <undefined>
2019-08-04 19:24:22.348671: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcuda.so.1
2019-08-04 19:24:22.573397: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:24:22.574025: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x95121c0 executing computations on platform CUDA. Devices:
2019-08-04 19:24:22.574064: I tensorflow/compiler/xla/service/service.cc:175]   StreamExecutor device (0): Tesla T4, Compute Capability 7.5
2019-08-04 19:24:22.574401: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:24:22.575001: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 19:24:22.588746: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 19:24:22.758388: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 19:24:22.838309: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 19:24:22.863166: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 19:24:23.062855: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 19:24:23.170580: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 19:24:23.517111: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 19:24:23.517378: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:24:23.517903: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:24:23.518245: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 19:24:23.520902: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 19:24:23.522597: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 19:24:23.522625: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 19:24:23.522643: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 19:24:23.525086: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:24:23.525589: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:24:23.525942: W tensorflow/core/common_runtime/gpu/gpu_bfc_allocator.cc:40] Overriding allow_growth setting because the TF_FORCE_GPU_ALLOW_GROWTH environment variable is set. Original config value was 0.
2019-08-04 19:24:23.525985: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
2019-08-04 19:24:23.873368: W tensorflow/compiler/jit/mark_for_compilation_pass.cc:1412] (One-time warning): Not using XLA:CPU for cluster because envvar TF_XLA_FLAGS=--tf_xla_cpu_global_jit was not set.  If you want XLA:CPU, either set that envvar, or use experimental_jit_scope to enable XLA:CPU.  To confirm that XLA is active, pass --vmodule=xla_compilation_cache=1 (as a proper command-line flag, not via TF_XLA_FLAGS) or set the envvar XLA_FLAGS=--xla_hlo_profile.
I0804 19:24:25.580031 140200711067520 session_manager.py:500] Running local_init_op.
I0804 19:24:25.614005 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 19:24:27.651640 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 0 into experiment/transformer/transformer_small/output/model.ckpt.
2019-08-04 19:24:29.339574: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
I0804 19:24:32.483394 140200711067520 basic_session_run_hooks.py:262] loss = 8.155501, step = 0
I0804 19:24:36.518196 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 24.7793
I0804 19:24:36.519674 140200711067520 basic_session_run_hooks.py:260] loss = 5.2857046, step = 100 (4.036 sec)
I0804 19:24:39.498749 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.551
I0804 19:24:39.500139 140200711067520 basic_session_run_hooks.py:260] loss = 3.4506884, step = 200 (2.980 sec)
I0804 19:24:42.455612 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.8195
I0804 19:24:42.457046 140200711067520 basic_session_run_hooks.py:260] loss = 3.0842905, step = 300 (2.957 sec)
I0804 19:24:45.409548 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.8535
I0804 19:24:45.411277 140200711067520 basic_session_run_hooks.py:260] loss = 3.058373, step = 400 (2.954 sec)
I0804 19:24:48.360075 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.8921
I0804 19:24:48.361675 140200711067520 basic_session_run_hooks.py:260] loss = 2.8275805, step = 500 (2.950 sec)
I0804 19:24:51.305898 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.9463
I0804 19:24:51.307310 140200711067520 basic_session_run_hooks.py:260] loss = 2.579091, step = 600 (2.946 sec)
I0804 19:24:54.257405 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.8808
I0804 19:24:54.258950 140200711067520 basic_session_run_hooks.py:260] loss = 2.4020169, step = 700 (2.952 sec)
I0804 19:24:57.242395 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.501
I0804 19:24:57.243805 140200711067520 basic_session_run_hooks.py:260] loss = 2.4527538, step = 800 (2.985 sec)
I0804 19:25:00.233128 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.4369
I0804 19:25:00.234609 140200711067520 basic_session_run_hooks.py:260] loss = 2.389892, step = 900 (2.991 sec)
I0804 19:25:03.196714 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 1000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:25:03.523173 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 19:25:03.524591 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 19:25:03.679222 140200711067520 estimator.py:1145] Calling model_fn.
I0804 19:25:03.680652 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 19:25:03.681203 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 19:25:03.681339 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 19:25:03.681461 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 19:25:03.681557 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 19:25:03.681661 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 19:25:03.681751 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 19:25:03.776337 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 19:25:03.840514 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 19:25:03.988373 140200711067520 t2t_model.py:2172] Building model body
I0804 19:25:04.936924 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
W0804 19:25:05.102966 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/metrics.py:582: The name tf.metrics.mean is deprecated. Please use tf.compat.v1.metrics.mean instead.

W0804 19:25:05.505138 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/bleu_hook.py:151: py_func (from tensorflow.python.ops.script_ops) is deprecated and will be removed in a future version.
Instructions for updating:
tf.py_func is deprecated in TF V2. Instead, there are two
    options available in V2.
    - tf.py_function takes a python function which manipulates tf eager
    tensors instead of numpy arrays. It's easy to convert a tf eager tensor to
    an ndarray (just call tensor.numpy()) but having access to eager tensors
    means `tf.py_function`s can use accelerators such as GPUs as well as
    being differentiable using a gradient tape.
    - tf.numpy_function maintains the semantics of the deprecated tf.py_func
    (it is not differentiable, and manipulates numpy arrays). It drops the
    stateful argument making all functions stateful.
    
W0804 19:25:05.688900 140200711067520 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/tensor2tensor/utils/t2t_model.py:1670: The name tf.summary.merge_all is deprecated. Please use tf.compat.v1.summary.merge_all instead.

I0804 19:25:05.690661 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 19:25:05.713266 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T19:25:05Z
I0804 19:25:06.189876 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 19:25:06.190700: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:25:06.191100: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 19:25:06.191209: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 19:25:06.191234: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 19:25:06.191259: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 19:25:06.191281: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 19:25:06.191306: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 19:25:06.191328: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 19:25:06.191352: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 19:25:06.191497: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:25:06.191925: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:25:06.192297: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 19:25:06.192339: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 19:25:06.192354: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 19:25:06.192365: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 19:25:06.192703: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:25:06.193130: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:25:06.193514: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
W0804 19:25:06.193651 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/training/saver.py:1276: checkpoint_exists (from tensorflow.python.training.checkpoint_management) is deprecated and will be removed in a future version.
Instructions for updating:
Use standard file APIs to check for files with this prefix.
I0804 19:25:06.194862 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-1000
I0804 19:25:06.383098 140200711067520 session_manager.py:500] Running local_init_op.
I0804 19:25:06.426431 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 19:25:12.864027 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 19:25:18.602217 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 19:25:24.269555 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 19:25:30.000144 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 19:25:35.615264 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 19:25:41.510267 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 19:25:47.200354 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 19:25:52.884746 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 19:25:58.551951 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 19:26:03.668064 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-19:26:03
I0804 19:26:03.668334 140200711067520 estimator.py:2039] Saving dict for global step 1000: global_step = 1000, loss = 2.7128453, metrics-paper_generation_problem/targets/accuracy = 0.26113552, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.6263957, metrics-paper_generation_problem/targets/approx_bleu_score = 0.12949093, metrics-paper_generation_problem/targets/neg_log_perplexity = -2.712865, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.25634146, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.3960722
I0804 19:26:03.668929 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 1000: experiment/transformer/transformer_small/output/model.ckpt-1000
I0804 19:26:03.720886 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.57511
I0804 19:26:03.722041 140200711067520 basic_session_run_hooks.py:260] loss = 2.409999, step = 1000 (63.487 sec)
I0804 19:26:06.732197 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.2086
I0804 19:26:06.733334 140200711067520 basic_session_run_hooks.py:260] loss = 2.379183, step = 1100 (3.011 sec)
I0804 19:26:09.694392 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.7588
I0804 19:26:09.695862 140200711067520 basic_session_run_hooks.py:260] loss = 2.2706354, step = 1200 (2.963 sec)
I0804 19:26:12.664976 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.6638
I0804 19:26:12.666488 140200711067520 basic_session_run_hooks.py:260] loss = 2.1597714, step = 1300 (2.971 sec)
I0804 19:26:15.605208 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 34.0104
I0804 19:26:15.606709 140200711067520 basic_session_run_hooks.py:260] loss = 2.1520514, step = 1400 (2.940 sec)
I0804 19:26:18.582639 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.5867
I0804 19:26:18.584065 140200711067520 basic_session_run_hooks.py:260] loss = 2.2127125, step = 1500 (2.977 sec)
I0804 19:26:21.566531 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.5128
I0804 19:26:21.567976 140200711067520 basic_session_run_hooks.py:260] loss = 2.0944536, step = 1600 (2.984 sec)
I0804 19:26:24.546823 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.5536
I0804 19:26:24.548582 140200711067520 basic_session_run_hooks.py:260] loss = 2.0609598, step = 1700 (2.981 sec)
I0804 19:26:27.561280 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.1735
I0804 19:26:27.562687 140200711067520 basic_session_run_hooks.py:260] loss = 1.9879534, step = 1800 (3.014 sec)
I0804 19:26:30.551955 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.4397
I0804 19:26:30.554009 140200711067520 basic_session_run_hooks.py:260] loss = 1.8525437, step = 1900 (2.991 sec)
I0804 19:26:33.533509 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 2000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:26:33.856554 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:26:33.890398 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.952
I0804 19:26:33.891387 140200711067520 basic_session_run_hooks.py:260] loss = 1.8864993, step = 2000 (3.337 sec)
I0804 19:26:36.900977 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.2162
I0804 19:26:36.902472 140200711067520 basic_session_run_hooks.py:260] loss = 1.8693491, step = 2100 (3.011 sec)
I0804 19:26:39.917082 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.1554
I0804 19:26:39.918612 140200711067520 basic_session_run_hooks.py:260] loss = 1.7641883, step = 2200 (3.016 sec)
I0804 19:26:42.932776 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.16
I0804 19:26:42.933911 140200711067520 basic_session_run_hooks.py:260] loss = 1.7817109, step = 2300 (3.015 sec)
I0804 19:26:45.979623 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8208
I0804 19:26:45.980860 140200711067520 basic_session_run_hooks.py:260] loss = 1.6913946, step = 2400 (3.047 sec)
I0804 19:26:49.032759 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7533
I0804 19:26:49.034281 140200711067520 basic_session_run_hooks.py:260] loss = 1.7186183, step = 2500 (3.053 sec)
I0804 19:26:52.098445 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.619
I0804 19:26:52.099894 140200711067520 basic_session_run_hooks.py:260] loss = 1.6929767, step = 2600 (3.066 sec)
I0804 19:26:55.139570 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8828
I0804 19:26:55.141005 140200711067520 basic_session_run_hooks.py:260] loss = 1.5994068, step = 2700 (3.041 sec)
I0804 19:26:58.194659 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7319
I0804 19:26:58.196120 140200711067520 basic_session_run_hooks.py:260] loss = 1.6039113, step = 2800 (3.055 sec)
I0804 19:27:01.242408 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8113
I0804 19:27:01.243752 140200711067520 basic_session_run_hooks.py:260] loss = 1.6295178, step = 2900 (3.048 sec)
I0804 19:27:04.237113 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 3000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:27:04.550706 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:27:04.586644 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.9019
I0804 19:27:04.587828 140200711067520 basic_session_run_hooks.py:260] loss = 1.6565264, step = 3000 (3.344 sec)
I0804 19:27:07.640507 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.746
I0804 19:27:07.641699 140200711067520 basic_session_run_hooks.py:260] loss = 1.5527774, step = 3100 (3.054 sec)
I0804 19:27:10.707409 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6059
I0804 19:27:10.708856 140200711067520 basic_session_run_hooks.py:260] loss = 1.5831912, step = 3200 (3.067 sec)
I0804 19:27:13.774051 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6089
I0804 19:27:13.775332 140200711067520 basic_session_run_hooks.py:260] loss = 1.6109092, step = 3300 (3.066 sec)
I0804 19:27:16.842576 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5891
I0804 19:27:16.843778 140200711067520 basic_session_run_hooks.py:260] loss = 1.4197326, step = 3400 (3.068 sec)
I0804 19:27:19.892964 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7826
I0804 19:27:19.894052 140200711067520 basic_session_run_hooks.py:260] loss = 1.5294964, step = 3500 (3.050 sec)
I0804 19:27:22.921457 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0201
I0804 19:27:22.923023 140200711067520 basic_session_run_hooks.py:260] loss = 1.5571033, step = 3600 (3.029 sec)
I0804 19:27:25.939731 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.1311
I0804 19:27:25.941386 140200711067520 basic_session_run_hooks.py:260] loss = 1.4533595, step = 3700 (3.018 sec)
I0804 19:27:28.964576 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0598
I0804 19:27:28.965959 140200711067520 basic_session_run_hooks.py:260] loss = 1.5444443, step = 3800 (3.025 sec)
I0804 19:27:31.960324 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.3803
I0804 19:27:31.961987 140200711067520 basic_session_run_hooks.py:260] loss = 1.518613, step = 3900 (2.996 sec)
I0804 19:27:34.944219 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 4000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:27:35.268598 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:27:35.304935 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.8987
I0804 19:27:35.306002 140200711067520 basic_session_run_hooks.py:260] loss = 1.4616487, step = 4000 (3.344 sec)
I0804 19:27:38.312713 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.2474
I0804 19:27:38.313874 140200711067520 basic_session_run_hooks.py:260] loss = 1.507455, step = 4100 (3.008 sec)
I0804 19:27:41.351313 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9102
I0804 19:27:41.353080 140200711067520 basic_session_run_hooks.py:260] loss = 1.4154464, step = 4200 (3.039 sec)
I0804 19:27:44.383909 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.975
I0804 19:27:44.385295 140200711067520 basic_session_run_hooks.py:260] loss = 1.5419605, step = 4300 (3.032 sec)
I0804 19:27:47.412206 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0216
I0804 19:27:47.413871 140200711067520 basic_session_run_hooks.py:260] loss = 1.4361955, step = 4400 (3.029 sec)
I0804 19:27:50.440459 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0227
I0804 19:27:50.442209 140200711067520 basic_session_run_hooks.py:260] loss = 1.3980746, step = 4500 (3.028 sec)
I0804 19:27:53.486541 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8291
I0804 19:27:53.488100 140200711067520 basic_session_run_hooks.py:260] loss = 1.4099467, step = 4600 (3.046 sec)
I0804 19:27:56.535737 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7951
I0804 19:27:56.537252 140200711067520 basic_session_run_hooks.py:260] loss = 1.4482095, step = 4700 (3.049 sec)
I0804 19:27:59.606945 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5603
I0804 19:27:59.608086 140200711067520 basic_session_run_hooks.py:260] loss = 1.5307233, step = 4800 (3.071 sec)
I0804 19:28:02.622036 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.1666
I0804 19:28:02.623761 140200711067520 basic_session_run_hooks.py:260] loss = 1.4048406, step = 4900 (3.016 sec)
I0804 19:28:05.611516 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 5000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:28:05.921041 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:28:05.960367 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.9551
I0804 19:28:05.961374 140200711067520 basic_session_run_hooks.py:260] loss = 1.4705069, step = 5000 (3.338 sec)
I0804 19:28:08.978475 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.1339
I0804 19:28:08.979931 140200711067520 basic_session_run_hooks.py:260] loss = 1.4336276, step = 5100 (3.019 sec)
I0804 19:28:12.036247 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7031
I0804 19:28:12.037760 140200711067520 basic_session_run_hooks.py:260] loss = 1.4010735, step = 5200 (3.058 sec)
I0804 19:28:15.064410 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0234
I0804 19:28:15.065931 140200711067520 basic_session_run_hooks.py:260] loss = 1.415313, step = 5300 (3.028 sec)
I0804 19:28:18.098437 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9596
I0804 19:28:18.099949 140200711067520 basic_session_run_hooks.py:260] loss = 1.4335316, step = 5400 (3.034 sec)
I0804 19:28:21.160341 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6592
I0804 19:28:21.162182 140200711067520 basic_session_run_hooks.py:260] loss = 1.3694171, step = 5500 (3.062 sec)
I0804 19:28:24.180371 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.1123
I0804 19:28:24.181669 140200711067520 basic_session_run_hooks.py:260] loss = 1.2803149, step = 5600 (3.019 sec)
I0804 19:28:27.215081 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9522
I0804 19:28:27.216775 140200711067520 basic_session_run_hooks.py:260] loss = 1.4415474, step = 5700 (3.035 sec)
I0804 19:28:30.263624 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8027
I0804 19:28:30.264824 140200711067520 basic_session_run_hooks.py:260] loss = 1.352975, step = 5800 (3.048 sec)
I0804 19:28:33.287676 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0682
I0804 19:28:33.289037 140200711067520 basic_session_run_hooks.py:260] loss = 1.3370267, step = 5900 (3.024 sec)
I0804 19:28:36.283484 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 6000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:28:36.577485 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:28:36.610884 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.091
I0804 19:28:36.611997 140200711067520 basic_session_run_hooks.py:260] loss = 1.3969012, step = 6000 (3.323 sec)
I0804 19:28:39.640194 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0115
I0804 19:28:39.641671 140200711067520 basic_session_run_hooks.py:260] loss = 1.3341318, step = 6100 (3.030 sec)
I0804 19:28:42.691244 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7753
I0804 19:28:42.692683 140200711067520 basic_session_run_hooks.py:260] loss = 1.4407762, step = 6200 (3.051 sec)
I0804 19:28:45.721908 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9962
I0804 19:28:45.723112 140200711067520 basic_session_run_hooks.py:260] loss = 1.3321823, step = 6300 (3.030 sec)
I0804 19:28:48.742534 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.1058
I0804 19:28:48.743758 140200711067520 basic_session_run_hooks.py:260] loss = 1.5399805, step = 6400 (3.021 sec)
I0804 19:28:51.820149 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4923
I0804 19:28:51.821729 140200711067520 basic_session_run_hooks.py:260] loss = 1.3282262, step = 6500 (3.078 sec)
I0804 19:28:54.839002 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.1257
I0804 19:28:54.840555 140200711067520 basic_session_run_hooks.py:260] loss = 1.2764181, step = 6600 (3.019 sec)
I0804 19:28:57.865550 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0407
I0804 19:28:57.867305 140200711067520 basic_session_run_hooks.py:260] loss = 1.3656299, step = 6700 (3.027 sec)
I0804 19:29:00.895458 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0043
I0804 19:29:00.896995 140200711067520 basic_session_run_hooks.py:260] loss = 1.3575528, step = 6800 (3.030 sec)
I0804 19:29:03.941338 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8309
I0804 19:29:03.942813 140200711067520 basic_session_run_hooks.py:260] loss = 1.3987968, step = 6900 (3.046 sec)
I0804 19:29:06.971069 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 7000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:29:07.254149 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:29:07.295687 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.8119
I0804 19:29:07.296850 140200711067520 basic_session_run_hooks.py:260] loss = 1.4891579, step = 7000 (3.354 sec)
I0804 19:29:10.359506 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6394
I0804 19:29:10.361052 140200711067520 basic_session_run_hooks.py:260] loss = 1.2564094, step = 7100 (3.064 sec)
I0804 19:29:13.477219 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0746
I0804 19:29:13.478415 140200711067520 basic_session_run_hooks.py:260] loss = 1.3091213, step = 7200 (3.117 sec)
I0804 19:29:16.549064 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5538
I0804 19:29:16.550181 140200711067520 basic_session_run_hooks.py:260] loss = 1.3494519, step = 7300 (3.072 sec)
I0804 19:29:19.606631 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7057
I0804 19:29:19.607827 140200711067520 basic_session_run_hooks.py:260] loss = 1.1798173, step = 7400 (3.058 sec)
I0804 19:29:22.651954 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8372
I0804 19:29:22.653059 140200711067520 basic_session_run_hooks.py:260] loss = 1.1869992, step = 7500 (3.045 sec)
I0804 19:29:25.736833 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4166
I0804 19:29:25.738704 140200711067520 basic_session_run_hooks.py:260] loss = 1.3844441, step = 7600 (3.086 sec)
I0804 19:29:28.803191 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6115
I0804 19:29:28.804862 140200711067520 basic_session_run_hooks.py:260] loss = 1.3332542, step = 7700 (3.066 sec)
I0804 19:29:31.877873 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5237
I0804 19:29:31.879033 140200711067520 basic_session_run_hooks.py:260] loss = 1.294771, step = 7800 (3.074 sec)
I0804 19:29:34.926853 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7979
I0804 19:29:34.928403 140200711067520 basic_session_run_hooks.py:260] loss = 1.2601916, step = 7900 (3.049 sec)
I0804 19:29:37.965266 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 8000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:29:38.244294 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:29:38.277575 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.8443
I0804 19:29:38.278681 140200711067520 basic_session_run_hooks.py:260] loss = 1.3717446, step = 8000 (3.350 sec)
I0804 19:29:41.355464 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4899
I0804 19:29:41.356830 140200711067520 basic_session_run_hooks.py:260] loss = 1.3131135, step = 8100 (3.078 sec)
I0804 19:29:44.388844 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9664
I0804 19:29:44.390315 140200711067520 basic_session_run_hooks.py:260] loss = 1.2771983, step = 8200 (3.033 sec)
I0804 19:29:47.432667 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8535
I0804 19:29:47.434068 140200711067520 basic_session_run_hooks.py:260] loss = 1.3089756, step = 8300 (3.044 sec)
I0804 19:29:50.496766 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6359
I0804 19:29:50.498517 140200711067520 basic_session_run_hooks.py:260] loss = 1.3441107, step = 8400 (3.064 sec)
I0804 19:29:53.582629 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4058
I0804 19:29:53.584056 140200711067520 basic_session_run_hooks.py:260] loss = 1.2427619, step = 8500 (3.086 sec)
I0804 19:29:56.699201 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0872
I0804 19:29:56.700674 140200711067520 basic_session_run_hooks.py:260] loss = 1.2500312, step = 8600 (3.117 sec)
I0804 19:29:59.801394 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2347
I0804 19:29:59.802819 140200711067520 basic_session_run_hooks.py:260] loss = 1.2316062, step = 8700 (3.102 sec)
I0804 19:30:02.929850 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9647
I0804 19:30:02.931274 140200711067520 basic_session_run_hooks.py:260] loss = 1.1864159, step = 8800 (3.128 sec)
I0804 19:30:06.015702 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.406
I0804 19:30:06.017162 140200711067520 basic_session_run_hooks.py:260] loss = 1.3271396, step = 8900 (3.086 sec)
I0804 19:30:09.002612 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 9000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:30:09.280321 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:30:09.314185 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.3167
I0804 19:30:09.315178 140200711067520 basic_session_run_hooks.py:260] loss = 1.2230386, step = 9000 (3.298 sec)
I0804 19:30:12.411708 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2843
I0804 19:30:12.413269 140200711067520 basic_session_run_hooks.py:260] loss = 1.360263, step = 9100 (3.098 sec)
I0804 19:30:15.491156 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4733
I0804 19:30:15.492901 140200711067520 basic_session_run_hooks.py:260] loss = 1.3862758, step = 9200 (3.080 sec)
I0804 19:30:18.592680 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2422
I0804 19:30:18.593923 140200711067520 basic_session_run_hooks.py:260] loss = 1.2016011, step = 9300 (3.101 sec)
I0804 19:30:21.622977 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0001
I0804 19:30:21.624153 140200711067520 basic_session_run_hooks.py:260] loss = 1.2682439, step = 9400 (3.030 sec)
I0804 19:30:24.661991 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9054
I0804 19:30:24.663237 140200711067520 basic_session_run_hooks.py:260] loss = 1.3375763, step = 9500 (3.039 sec)
I0804 19:30:27.707390 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8367
I0804 19:30:27.709131 140200711067520 basic_session_run_hooks.py:260] loss = 1.2132281, step = 9600 (3.046 sec)
I0804 19:30:30.771943 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.631
I0804 19:30:30.773294 140200711067520 basic_session_run_hooks.py:260] loss = 1.338763, step = 9700 (3.064 sec)
I0804 19:30:33.862114 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3608
I0804 19:30:33.863699 140200711067520 basic_session_run_hooks.py:260] loss = 1.2979809, step = 9800 (3.090 sec)
I0804 19:30:36.965536 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2227
I0804 19:30:36.966934 140200711067520 basic_session_run_hooks.py:260] loss = 1.2382991, step = 9900 (3.103 sec)
I0804 19:30:40.036624 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 10000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:30:40.334321 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:30:40.372535 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.351
I0804 19:30:40.373526 140200711067520 basic_session_run_hooks.py:260] loss = 1.1605426, step = 10000 (3.407 sec)
I0804 19:30:43.472177 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.262
I0804 19:30:43.473607 140200711067520 basic_session_run_hooks.py:260] loss = 1.180176, step = 10100 (3.100 sec)
I0804 19:30:46.511767 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8992
I0804 19:30:46.513055 140200711067520 basic_session_run_hooks.py:260] loss = 1.3104894, step = 10200 (3.039 sec)
I0804 19:30:49.558764 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8192
I0804 19:30:49.560128 140200711067520 basic_session_run_hooks.py:260] loss = 1.3073193, step = 10300 (3.047 sec)
I0804 19:30:52.609038 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7838
I0804 19:30:52.610279 140200711067520 basic_session_run_hooks.py:260] loss = 1.2635038, step = 10400 (3.050 sec)
I0804 19:30:55.678632 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5775
I0804 19:30:55.679919 140200711067520 basic_session_run_hooks.py:260] loss = 1.3505212, step = 10500 (3.070 sec)
I0804 19:30:58.745019 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6118
I0804 19:30:58.746312 140200711067520 basic_session_run_hooks.py:260] loss = 1.2462518, step = 10600 (3.066 sec)
I0804 19:31:01.830933 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4053
I0804 19:31:01.832185 140200711067520 basic_session_run_hooks.py:260] loss = 1.2149142, step = 10700 (3.086 sec)
I0804 19:31:04.909625 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4814
I0804 19:31:04.910925 140200711067520 basic_session_run_hooks.py:260] loss = 1.2501847, step = 10800 (3.079 sec)
I0804 19:31:07.989155 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4725
I0804 19:31:07.990710 140200711067520 basic_session_run_hooks.py:260] loss = 1.2028272, step = 10900 (3.080 sec)
I0804 19:31:10.990052 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 11000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:31:11.271090 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:31:11.305317 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.155
I0804 19:31:11.306512 140200711067520 basic_session_run_hooks.py:260] loss = 1.2102648, step = 11000 (3.316 sec)
I0804 19:31:14.342254 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9283
I0804 19:31:14.343698 140200711067520 basic_session_run_hooks.py:260] loss = 1.2601262, step = 11100 (3.037 sec)
I0804 19:31:17.368982 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.039
I0804 19:31:17.370527 140200711067520 basic_session_run_hooks.py:260] loss = 1.3274779, step = 11200 (3.027 sec)
I0804 19:31:20.392978 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0688
I0804 19:31:20.394392 140200711067520 basic_session_run_hooks.py:260] loss = 1.1980853, step = 11300 (3.024 sec)
I0804 19:31:23.397733 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.2804
I0804 19:31:23.399140 140200711067520 basic_session_run_hooks.py:260] loss = 1.4138035, step = 11400 (3.005 sec)
I0804 19:31:26.441767 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8514
I0804 19:31:26.443464 140200711067520 basic_session_run_hooks.py:260] loss = 1.1718758, step = 11500 (3.044 sec)
I0804 19:31:29.444451 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.3036
I0804 19:31:29.445904 140200711067520 basic_session_run_hooks.py:260] loss = 1.227898, step = 11600 (3.002 sec)
I0804 19:31:32.510579 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6145
I0804 19:31:32.511914 140200711067520 basic_session_run_hooks.py:260] loss = 1.2353508, step = 11700 (3.066 sec)
I0804 19:31:35.530181 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.117
I0804 19:31:35.531721 140200711067520 basic_session_run_hooks.py:260] loss = 1.2210552, step = 11800 (3.020 sec)
I0804 19:31:38.590711 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6735
I0804 19:31:38.591968 140200711067520 basic_session_run_hooks.py:260] loss = 1.2374473, step = 11900 (3.060 sec)
I0804 19:31:41.607785 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 12000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:31:41.888094 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:31:41.929655 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.9495
I0804 19:31:41.931216 140200711067520 basic_session_run_hooks.py:260] loss = 1.2551806, step = 12000 (3.339 sec)
I0804 19:31:44.980729 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7756
I0804 19:31:44.981899 140200711067520 basic_session_run_hooks.py:260] loss = 1.159721, step = 12100 (3.051 sec)
I0804 19:31:48.044164 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.643
I0804 19:31:48.045541 140200711067520 basic_session_run_hooks.py:260] loss = 1.2198614, step = 12200 (3.064 sec)
I0804 19:31:51.111698 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5994
I0804 19:31:51.113060 140200711067520 basic_session_run_hooks.py:260] loss = 1.2674105, step = 12300 (3.068 sec)
I0804 19:31:54.192655 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4575
I0804 19:31:54.194410 140200711067520 basic_session_run_hooks.py:260] loss = 1.2431961, step = 12400 (3.081 sec)
I0804 19:31:57.306456 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1153
I0804 19:31:57.308202 140200711067520 basic_session_run_hooks.py:260] loss = 1.1959051, step = 12500 (3.114 sec)
I0804 19:32:00.363126 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7152
I0804 19:32:00.364523 140200711067520 basic_session_run_hooks.py:260] loss = 1.1595746, step = 12600 (3.056 sec)
I0804 19:32:03.410493 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8154
I0804 19:32:03.411881 140200711067520 basic_session_run_hooks.py:260] loss = 1.2487422, step = 12700 (3.047 sec)
I0804 19:32:06.477142 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6086
I0804 19:32:06.478483 140200711067520 basic_session_run_hooks.py:260] loss = 1.2127666, step = 12800 (3.067 sec)
I0804 19:32:09.558259 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4558
I0804 19:32:09.559898 140200711067520 basic_session_run_hooks.py:260] loss = 1.2425289, step = 12900 (3.081 sec)
I0804 19:32:12.591459 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 13000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:32:12.870286 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:32:12.910022 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.8349
I0804 19:32:12.911083 140200711067520 basic_session_run_hooks.py:260] loss = 1.2182281, step = 13000 (3.351 sec)
I0804 19:32:16.010016 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2583
I0804 19:32:16.011257 140200711067520 basic_session_run_hooks.py:260] loss = 1.1973971, step = 13100 (3.100 sec)
I0804 19:32:19.125058 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1023
I0804 19:32:19.126234 140200711067520 basic_session_run_hooks.py:260] loss = 1.1959229, step = 13200 (3.115 sec)
I0804 19:32:22.168682 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8555
I0804 19:32:22.169960 140200711067520 basic_session_run_hooks.py:260] loss = 1.2638351, step = 13300 (3.044 sec)
I0804 19:32:25.177658 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.234
I0804 19:32:25.179447 140200711067520 basic_session_run_hooks.py:260] loss = 1.2654928, step = 13400 (3.009 sec)
I0804 19:32:28.211886 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9573
I0804 19:32:28.213636 140200711067520 basic_session_run_hooks.py:260] loss = 1.1349357, step = 13500 (3.034 sec)
I0804 19:32:31.248620 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9301
I0804 19:32:31.249807 140200711067520 basic_session_run_hooks.py:260] loss = 1.1719645, step = 13600 (3.036 sec)
I0804 19:32:34.282307 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9634
I0804 19:32:34.283779 140200711067520 basic_session_run_hooks.py:260] loss = 1.2101492, step = 13700 (3.034 sec)
I0804 19:32:37.334640 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7616
I0804 19:32:37.335915 140200711067520 basic_session_run_hooks.py:260] loss = 1.1115232, step = 13800 (3.052 sec)
I0804 19:32:40.407629 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5416
I0804 19:32:40.408902 140200711067520 basic_session_run_hooks.py:260] loss = 1.1840647, step = 13900 (3.073 sec)
I0804 19:32:43.444034 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 14000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:32:43.723935 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:32:43.763708 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.7966
I0804 19:32:43.764799 140200711067520 basic_session_run_hooks.py:260] loss = 1.1738458, step = 14000 (3.356 sec)
I0804 19:32:46.862951 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2662
I0804 19:32:46.864213 140200711067520 basic_session_run_hooks.py:260] loss = 1.2190872, step = 14100 (3.099 sec)
I0804 19:32:49.952946 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3625
I0804 19:32:49.954496 140200711067520 basic_session_run_hooks.py:260] loss = 1.1474192, step = 14200 (3.090 sec)
I0804 19:32:53.020766 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5962
I0804 19:32:53.022028 140200711067520 basic_session_run_hooks.py:260] loss = 1.2996737, step = 14300 (3.068 sec)
I0804 19:32:56.078085 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7084
I0804 19:32:56.079457 140200711067520 basic_session_run_hooks.py:260] loss = 1.1575891, step = 14400 (3.057 sec)
I0804 19:32:59.134395 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7193
I0804 19:32:59.135835 140200711067520 basic_session_run_hooks.py:260] loss = 1.1813397, step = 14500 (3.056 sec)
I0804 19:33:02.225768 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.348
I0804 19:33:02.227131 140200711067520 basic_session_run_hooks.py:260] loss = 1.1509129, step = 14600 (3.091 sec)
I0804 19:33:05.300163 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5267
I0804 19:33:05.301403 140200711067520 basic_session_run_hooks.py:260] loss = 1.1458942, step = 14700 (3.074 sec)
I0804 19:33:08.390953 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3543
I0804 19:33:08.392508 140200711067520 basic_session_run_hooks.py:260] loss = 1.2474288, step = 14800 (3.091 sec)
I0804 19:33:11.492274 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2445
I0804 19:33:11.493993 140200711067520 basic_session_run_hooks.py:260] loss = 1.0441868, step = 14900 (3.101 sec)
I0804 19:33:14.511921 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 15000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:33:14.803041 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:33:14.842698 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.8466
I0804 19:33:14.843883 140200711067520 basic_session_run_hooks.py:260] loss = 1.1936878, step = 15000 (3.350 sec)
I0804 19:33:17.916839 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5296
I0804 19:33:17.918240 140200711067520 basic_session_run_hooks.py:260] loss = 1.2710189, step = 15100 (3.074 sec)
I0804 19:33:20.985048 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5924
I0804 19:33:20.986593 140200711067520 basic_session_run_hooks.py:260] loss = 1.2386189, step = 15200 (3.068 sec)
I0804 19:33:24.052024 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6054
I0804 19:33:24.053371 140200711067520 basic_session_run_hooks.py:260] loss = 1.1741124, step = 15300 (3.067 sec)
I0804 19:33:27.132543 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4623
I0804 19:33:27.133707 140200711067520 basic_session_run_hooks.py:260] loss = 1.2481456, step = 15400 (3.080 sec)
I0804 19:33:30.199351 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6069
I0804 19:33:30.200798 140200711067520 basic_session_run_hooks.py:260] loss = 1.0897235, step = 15500 (3.067 sec)
I0804 19:33:33.301364 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2373
I0804 19:33:33.302937 140200711067520 basic_session_run_hooks.py:260] loss = 1.1586254, step = 15600 (3.102 sec)
I0804 19:33:36.291994 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.4381
I0804 19:33:36.293476 140200711067520 basic_session_run_hooks.py:260] loss = 1.1332475, step = 15700 (2.991 sec)
I0804 19:33:39.290340 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.3513
I0804 19:33:39.291769 140200711067520 basic_session_run_hooks.py:260] loss = 1.1740195, step = 15800 (2.998 sec)
I0804 19:33:42.327117 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9296
I0804 19:33:42.328668 140200711067520 basic_session_run_hooks.py:260] loss = 1.2191907, step = 15900 (3.037 sec)
I0804 19:33:45.334685 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 16000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:33:45.609776 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:33:45.646860 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.1225
I0804 19:33:45.647949 140200711067520 basic_session_run_hooks.py:260] loss = 1.0809509, step = 16000 (3.319 sec)
I0804 19:33:48.711970 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6256
I0804 19:33:48.713227 140200711067520 basic_session_run_hooks.py:260] loss = 1.1982739, step = 16100 (3.065 sec)
I0804 19:33:51.773390 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6646
I0804 19:33:51.774976 140200711067520 basic_session_run_hooks.py:260] loss = 1.2547654, step = 16200 (3.062 sec)
I0804 19:33:54.828641 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7305
I0804 19:33:54.829794 140200711067520 basic_session_run_hooks.py:260] loss = 1.1052699, step = 16300 (3.055 sec)
I0804 19:33:57.903104 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.526
I0804 19:33:57.904279 140200711067520 basic_session_run_hooks.py:260] loss = 1.1795783, step = 16400 (3.074 sec)
I0804 19:34:00.935705 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.975
I0804 19:34:00.937056 140200711067520 basic_session_run_hooks.py:260] loss = 1.2374704, step = 16500 (3.033 sec)
I0804 19:34:03.978750 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8621
I0804 19:34:03.980371 140200711067520 basic_session_run_hooks.py:260] loss = 1.2353446, step = 16600 (3.043 sec)
I0804 19:34:07.039579 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6709
I0804 19:34:07.041054 140200711067520 basic_session_run_hooks.py:260] loss = 1.1759075, step = 16700 (3.061 sec)
I0804 19:34:10.104204 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6301
I0804 19:34:10.105698 140200711067520 basic_session_run_hooks.py:260] loss = 1.1354991, step = 16800 (3.065 sec)
I0804 19:34:13.183482 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4753
I0804 19:34:13.185091 140200711067520 basic_session_run_hooks.py:260] loss = 1.2012011, step = 16900 (3.079 sec)
I0804 19:34:16.225773 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 17000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:34:16.499884 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:34:16.540835 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.7851
I0804 19:34:16.541842 140200711067520 basic_session_run_hooks.py:260] loss = 1.1836874, step = 17000 (3.357 sec)
I0804 19:34:19.645821 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2066
I0804 19:34:19.647203 140200711067520 basic_session_run_hooks.py:260] loss = 1.1589793, step = 17100 (3.105 sec)
I0804 19:34:22.735124 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3697
I0804 19:34:22.736683 140200711067520 basic_session_run_hooks.py:260] loss = 1.1926513, step = 17200 (3.089 sec)
I0804 19:34:25.773269 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9149
I0804 19:34:25.774763 140200711067520 basic_session_run_hooks.py:260] loss = 1.1059518, step = 17300 (3.038 sec)
I0804 19:34:28.807503 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9573
I0804 19:34:28.809046 140200711067520 basic_session_run_hooks.py:260] loss = 1.2031003, step = 17400 (3.034 sec)
I0804 19:34:31.853115 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8339
I0804 19:34:31.854511 140200711067520 basic_session_run_hooks.py:260] loss = 1.1429232, step = 17500 (3.045 sec)
I0804 19:34:34.910079 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7122
I0804 19:34:34.911245 140200711067520 basic_session_run_hooks.py:260] loss = 1.212168, step = 17600 (3.057 sec)
I0804 19:34:37.976313 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6134
I0804 19:34:37.977970 140200711067520 basic_session_run_hooks.py:260] loss = 1.14036, step = 17700 (3.067 sec)
I0804 19:34:41.053347 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4988
I0804 19:34:41.054639 140200711067520 basic_session_run_hooks.py:260] loss = 1.207542, step = 17800 (3.077 sec)
I0804 19:34:44.123361 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.573
I0804 19:34:44.125077 140200711067520 basic_session_run_hooks.py:260] loss = 1.2870129, step = 17900 (3.070 sec)
I0804 19:34:47.147558 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 18000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:34:47.425818 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:34:47.465058 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.9248
I0804 19:34:47.466052 140200711067520 basic_session_run_hooks.py:260] loss = 1.1035069, step = 18000 (3.341 sec)
I0804 19:34:50.530331 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6239
I0804 19:34:50.531792 140200711067520 basic_session_run_hooks.py:260] loss = 1.1633078, step = 18100 (3.066 sec)
I0804 19:34:53.584231 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7448
I0804 19:34:53.585653 140200711067520 basic_session_run_hooks.py:260] loss = 1.1383585, step = 18200 (3.054 sec)
I0804 19:34:56.624778 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.889
I0804 19:34:56.626069 140200711067520 basic_session_run_hooks.py:260] loss = 1.200794, step = 18300 (3.040 sec)
I0804 19:34:59.670257 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8355
I0804 19:34:59.671993 140200711067520 basic_session_run_hooks.py:260] loss = 1.2173938, step = 18400 (3.046 sec)
I0804 19:35:02.712268 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.873
I0804 19:35:02.714134 140200711067520 basic_session_run_hooks.py:260] loss = 1.1468053, step = 18500 (3.042 sec)
I0804 19:35:05.741981 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0063
I0804 19:35:05.743324 140200711067520 basic_session_run_hooks.py:260] loss = 1.277636, step = 18600 (3.029 sec)
I0804 19:35:08.766356 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0646
I0804 19:35:08.767856 140200711067520 basic_session_run_hooks.py:260] loss = 1.1514399, step = 18700 (3.025 sec)
I0804 19:35:11.811201 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8426
I0804 19:35:11.812715 140200711067520 basic_session_run_hooks.py:260] loss = 1.1231668, step = 18800 (3.045 sec)
I0804 19:35:14.879994 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5861
I0804 19:35:14.881193 140200711067520 basic_session_run_hooks.py:260] loss = 1.1923566, step = 18900 (3.068 sec)
I0804 19:35:17.894748 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 19000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:35:18.178933 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 19:35:18.180367 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 19:35:18.331510 140200711067520 estimator.py:1145] Calling model_fn.
I0804 19:35:18.332573 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 19:35:18.332980 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 19:35:18.333073 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 19:35:18.333155 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 19:35:18.333221 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 19:35:18.333304 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 19:35:18.333377 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 19:35:18.421065 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 19:35:18.481137 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 19:35:18.619872 140200711067520 t2t_model.py:2172] Building model body
I0804 19:35:19.307488 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 19:35:20.314753 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 19:35:20.333621 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T19:35:20Z
I0804 19:35:20.501849 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 19:35:20.502474: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:35:20.502866: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 19:35:20.502975: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 19:35:20.503000: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 19:35:20.503025: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 19:35:20.503047: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 19:35:20.503068: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 19:35:20.503091: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 19:35:20.503115: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 19:35:20.503228: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:35:20.503694: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:35:20.504008: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 19:35:20.504050: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 19:35:20.504063: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 19:35:20.504074: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 19:35:20.504353: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:35:20.504758: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:35:20.505083: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 19:35:20.506659 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-19000
I0804 19:35:20.711789 140200711067520 session_manager.py:500] Running local_init_op.
I0804 19:35:20.759563 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 19:35:26.769286 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 19:35:32.076843 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 19:35:37.385871 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 19:35:42.723874 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 19:35:48.036636 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 19:35:53.423947 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 19:35:58.816497 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 19:36:04.211121 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 19:36:09.521650 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 19:36:14.369156 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-19:36:14
I0804 19:36:14.369442 140200711067520 estimator.py:2039] Saving dict for global step 19000: global_step = 19000, loss = 1.2819614, metrics-paper_generation_problem/targets/accuracy = 0.6460767, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8694033, metrics-paper_generation_problem/targets/approx_bleu_score = 0.4516523, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.2820032, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5532905, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.67241085
I0804 19:36:14.370100 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 19000: experiment/transformer/transformer_small/output/model.ckpt-19000
I0804 19:36:14.424113 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.67943
I0804 19:36:14.425105 140200711067520 basic_session_run_hooks.py:260] loss = 1.0774114, step = 19000 (59.544 sec)
I0804 19:36:17.545263 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.04
I0804 19:36:17.546634 140200711067520 basic_session_run_hooks.py:260] loss = 1.2865647, step = 19100 (3.122 sec)
I0804 19:36:20.643223 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2794
I0804 19:36:20.644990 140200711067520 basic_session_run_hooks.py:260] loss = 1.177287, step = 19200 (3.098 sec)
I0804 19:36:23.667823 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 33.0622
I0804 19:36:23.669291 140200711067520 basic_session_run_hooks.py:260] loss = 1.2077243, step = 19300 (3.024 sec)
I0804 19:36:26.740880 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5409
I0804 19:36:26.742123 140200711067520 basic_session_run_hooks.py:260] loss = 1.1630546, step = 19400 (3.073 sec)
I0804 19:36:29.852913 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1333
I0804 19:36:29.854687 140200711067520 basic_session_run_hooks.py:260] loss = 1.1744058, step = 19500 (3.113 sec)
I0804 19:36:32.968632 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0952
I0804 19:36:32.970390 140200711067520 basic_session_run_hooks.py:260] loss = 1.1371635, step = 19600 (3.116 sec)
I0804 19:36:36.103107 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9034
I0804 19:36:36.104585 140200711067520 basic_session_run_hooks.py:260] loss = 1.1755279, step = 19700 (3.134 sec)
I0804 19:36:39.256193 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7149
I0804 19:36:39.257328 140200711067520 basic_session_run_hooks.py:260] loss = 1.1414742, step = 19800 (3.153 sec)
I0804 19:36:42.421108 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5964
I0804 19:36:42.422265 140200711067520 basic_session_run_hooks.py:260] loss = 1.16572, step = 19900 (3.165 sec)
I0804 19:36:45.534163 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 20000 into experiment/transformer/transformer_small/output/model.ckpt.
W0804 19:36:45.561821 140200711067520 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/training/saver.py:960: remove_checkpoint (from tensorflow.python.training.checkpoint_management) is deprecated and will be removed in a future version.
Instructions for updating:
Use standard file APIs to delete files with this prefix.
I0804 19:36:45.824675 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:36:45.857781 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0977
I0804 19:36:45.858798 140200711067520 basic_session_run_hooks.py:260] loss = 1.1964328, step = 20000 (3.437 sec)
I0804 19:36:48.946603 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3753
I0804 19:36:48.947993 140200711067520 basic_session_run_hooks.py:260] loss = 1.2031322, step = 20100 (3.089 sec)
I0804 19:36:52.029267 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4391
I0804 19:36:52.030838 140200711067520 basic_session_run_hooks.py:260] loss = 1.1104207, step = 20200 (3.083 sec)
I0804 19:36:55.129518 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2557
I0804 19:36:55.131067 140200711067520 basic_session_run_hooks.py:260] loss = 1.1560361, step = 20300 (3.100 sec)
I0804 19:36:58.233558 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2161
I0804 19:36:58.235221 140200711067520 basic_session_run_hooks.py:260] loss = 1.2182827, step = 20400 (3.104 sec)
I0804 19:37:01.309588 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5094
I0804 19:37:01.310895 140200711067520 basic_session_run_hooks.py:260] loss = 1.1767564, step = 20500 (3.076 sec)
I0804 19:37:04.398446 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3744
I0804 19:37:04.399903 140200711067520 basic_session_run_hooks.py:260] loss = 1.1665094, step = 20600 (3.089 sec)
I0804 19:37:07.504444 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1957
I0804 19:37:07.506119 140200711067520 basic_session_run_hooks.py:260] loss = 1.0595956, step = 20700 (3.106 sec)
I0804 19:37:10.581554 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4981
I0804 19:37:10.583051 140200711067520 basic_session_run_hooks.py:260] loss = 1.1754962, step = 20800 (3.077 sec)
I0804 19:37:13.612342 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.9945
I0804 19:37:13.613761 140200711067520 basic_session_run_hooks.py:260] loss = 1.142103, step = 20900 (3.031 sec)
I0804 19:37:16.630002 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 21000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:37:16.931900 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:37:16.965882 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.819
I0804 19:37:16.966959 140200711067520 basic_session_run_hooks.py:260] loss = 1.1589005, step = 21000 (3.353 sec)
I0804 19:37:20.068833 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2276
I0804 19:37:20.070288 140200711067520 basic_session_run_hooks.py:260] loss = 1.0709862, step = 21100 (3.103 sec)
I0804 19:37:23.135121 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6128
I0804 19:37:23.136287 140200711067520 basic_session_run_hooks.py:260] loss = 1.1429499, step = 21200 (3.066 sec)
I0804 19:37:26.225683 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3565
I0804 19:37:26.227053 140200711067520 basic_session_run_hooks.py:260] loss = 1.214007, step = 21300 (3.091 sec)
I0804 19:37:29.324099 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2746
I0804 19:37:29.325542 140200711067520 basic_session_run_hooks.py:260] loss = 1.1842749, step = 21400 (3.098 sec)
I0804 19:37:32.401624 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4936
I0804 19:37:32.403098 140200711067520 basic_session_run_hooks.py:260] loss = 1.1561172, step = 21500 (3.078 sec)
I0804 19:37:35.475774 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5293
I0804 19:37:35.477524 140200711067520 basic_session_run_hooks.py:260] loss = 1.2352748, step = 21600 (3.074 sec)
I0804 19:37:38.517554 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8757
I0804 19:37:38.518861 140200711067520 basic_session_run_hooks.py:260] loss = 1.18102, step = 21700 (3.041 sec)
I0804 19:37:41.585565 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5947
I0804 19:37:41.586823 140200711067520 basic_session_run_hooks.py:260] loss = 1.1425862, step = 21800 (3.068 sec)
I0804 19:37:44.673245 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3862
I0804 19:37:44.674494 140200711067520 basic_session_run_hooks.py:260] loss = 1.2044507, step = 21900 (3.088 sec)
I0804 19:37:47.703058 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 22000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:37:47.998501 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:37:48.041297 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.6907
I0804 19:37:48.042533 140200711067520 basic_session_run_hooks.py:260] loss = 1.1608398, step = 22000 (3.368 sec)
I0804 19:37:51.162552 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0387
I0804 19:37:51.164051 140200711067520 basic_session_run_hooks.py:260] loss = 1.1650263, step = 22100 (3.122 sec)
I0804 19:37:54.292370 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9506
I0804 19:37:54.293650 140200711067520 basic_session_run_hooks.py:260] loss = 1.1885363, step = 22200 (3.130 sec)
I0804 19:37:57.417355 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0004
I0804 19:37:57.418799 140200711067520 basic_session_run_hooks.py:260] loss = 1.185166, step = 22300 (3.125 sec)
I0804 19:38:00.482517 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6247
I0804 19:38:00.484014 140200711067520 basic_session_run_hooks.py:260] loss = 1.1100837, step = 22400 (3.065 sec)
I0804 19:38:03.524507 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8732
I0804 19:38:03.525793 140200711067520 basic_session_run_hooks.py:260] loss = 1.2033346, step = 22500 (3.042 sec)
I0804 19:38:06.566694 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8708
I0804 19:38:06.568071 140200711067520 basic_session_run_hooks.py:260] loss = 1.142504, step = 22600 (3.042 sec)
I0804 19:38:09.629569 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6496
I0804 19:38:09.631026 140200711067520 basic_session_run_hooks.py:260] loss = 1.1963096, step = 22700 (3.063 sec)
I0804 19:38:12.732398 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2283
I0804 19:38:12.733744 140200711067520 basic_session_run_hooks.py:260] loss = 1.2043911, step = 22800 (3.103 sec)
I0804 19:38:15.832314 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.259
I0804 19:38:15.833613 140200711067520 basic_session_run_hooks.py:260] loss = 1.2078202, step = 22900 (3.100 sec)
I0804 19:38:18.877968 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 23000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:38:19.182590 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:38:19.225291 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.4723
I0804 19:38:19.226340 140200711067520 basic_session_run_hooks.py:260] loss = 1.0467908, step = 23000 (3.393 sec)
I0804 19:38:22.314800 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3679
I0804 19:38:22.316230 140200711067520 basic_session_run_hooks.py:260] loss = 1.1481651, step = 23100 (3.090 sec)
I0804 19:38:25.392672 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.49
I0804 19:38:25.394031 140200711067520 basic_session_run_hooks.py:260] loss = 1.2228857, step = 23200 (3.078 sec)
I0804 19:38:28.470298 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4927
I0804 19:38:28.471825 140200711067520 basic_session_run_hooks.py:260] loss = 1.1864159, step = 23300 (3.078 sec)
I0804 19:38:31.560724 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3578
I0804 19:38:31.562395 140200711067520 basic_session_run_hooks.py:260] loss = 1.1314461, step = 23400 (3.091 sec)
I0804 19:38:34.628766 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5941
I0804 19:38:34.630047 140200711067520 basic_session_run_hooks.py:260] loss = 1.1422839, step = 23500 (3.068 sec)
I0804 19:38:37.723471 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3135
I0804 19:38:37.724975 140200711067520 basic_session_run_hooks.py:260] loss = 1.0633789, step = 23600 (3.095 sec)
I0804 19:38:40.854739 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9358
I0804 19:38:40.855888 140200711067520 basic_session_run_hooks.py:260] loss = 1.1601166, step = 23700 (3.131 sec)
I0804 19:38:43.948783 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3203
I0804 19:38:43.950376 140200711067520 basic_session_run_hooks.py:260] loss = 1.1460757, step = 23800 (3.094 sec)
I0804 19:38:47.045925 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2877
I0804 19:38:47.047343 140200711067520 basic_session_run_hooks.py:260] loss = 1.0903977, step = 23900 (3.097 sec)
I0804 19:38:50.107812 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 24000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:38:50.392560 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:38:50.429239 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.5566
I0804 19:38:50.430280 140200711067520 basic_session_run_hooks.py:260] loss = 1.1522293, step = 24000 (3.383 sec)
I0804 19:38:53.500026 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5653
I0804 19:38:53.501578 140200711067520 basic_session_run_hooks.py:260] loss = 1.0931782, step = 24100 (3.071 sec)
I0804 19:38:56.591674 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3451
I0804 19:38:56.593028 140200711067520 basic_session_run_hooks.py:260] loss = 1.1558337, step = 24200 (3.091 sec)
I0804 19:38:59.696580 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2073
I0804 19:38:59.698130 140200711067520 basic_session_run_hooks.py:260] loss = 1.1288643, step = 24300 (3.105 sec)
I0804 19:39:02.803099 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1903
I0804 19:39:02.804646 140200711067520 basic_session_run_hooks.py:260] loss = 1.2165997, step = 24400 (3.107 sec)
I0804 19:39:05.914798 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1367
I0804 19:39:05.916656 140200711067520 basic_session_run_hooks.py:260] loss = 1.1749858, step = 24500 (3.112 sec)
I0804 19:39:09.025446 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1478
I0804 19:39:09.026866 140200711067520 basic_session_run_hooks.py:260] loss = 1.2127006, step = 24600 (3.110 sec)
I0804 19:39:12.174772 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7528
I0804 19:39:12.176238 140200711067520 basic_session_run_hooks.py:260] loss = 1.086857, step = 24700 (3.149 sec)
I0804 19:39:15.254091 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4747
I0804 19:39:15.255303 140200711067520 basic_session_run_hooks.py:260] loss = 1.1870553, step = 24800 (3.079 sec)
I0804 19:39:18.315655 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6631
I0804 19:39:18.317138 140200711067520 basic_session_run_hooks.py:260] loss = 1.2003901, step = 24900 (3.062 sec)
I0804 19:39:21.360117 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 25000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:39:21.663019 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:39:21.697321 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.5709
I0804 19:39:21.698580 140200711067520 basic_session_run_hooks.py:260] loss = 1.290784, step = 25000 (3.381 sec)
I0804 19:39:24.784930 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3877
I0804 19:39:24.786276 140200711067520 basic_session_run_hooks.py:260] loss = 1.1820152, step = 25100 (3.088 sec)
I0804 19:39:27.891233 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1927
I0804 19:39:27.892799 140200711067520 basic_session_run_hooks.py:260] loss = 1.0839361, step = 25200 (3.107 sec)
I0804 19:39:30.992761 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2421
I0804 19:39:30.994261 140200711067520 basic_session_run_hooks.py:260] loss = 1.2427635, step = 25300 (3.101 sec)
I0804 19:39:34.075691 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4366
I0804 19:39:34.077037 140200711067520 basic_session_run_hooks.py:260] loss = 1.2150134, step = 25400 (3.083 sec)
I0804 19:39:37.195538 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0532
I0804 19:39:37.197069 140200711067520 basic_session_run_hooks.py:260] loss = 1.150935, step = 25500 (3.120 sec)
I0804 19:39:40.262788 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6024
I0804 19:39:40.264772 140200711067520 basic_session_run_hooks.py:260] loss = 1.1871003, step = 25600 (3.068 sec)
I0804 19:39:43.327203 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6325
I0804 19:39:43.328678 140200711067520 basic_session_run_hooks.py:260] loss = 1.1857219, step = 25700 (3.064 sec)
I0804 19:39:46.411812 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.419
I0804 19:39:46.413258 140200711067520 basic_session_run_hooks.py:260] loss = 1.0898191, step = 25800 (3.085 sec)
I0804 19:39:49.502558 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3549
I0804 19:39:49.504561 140200711067520 basic_session_run_hooks.py:260] loss = 1.2438653, step = 25900 (3.091 sec)
I0804 19:39:52.550063 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 26000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:39:52.839120 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:39:52.879206 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.6148
I0804 19:39:52.880406 140200711067520 basic_session_run_hooks.py:260] loss = 1.1750416, step = 26000 (3.376 sec)
I0804 19:39:55.994803 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.097
I0804 19:39:55.996100 140200711067520 basic_session_run_hooks.py:260] loss = 1.1264752, step = 26100 (3.116 sec)
I0804 19:39:59.091130 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2962
I0804 19:39:59.092327 140200711067520 basic_session_run_hooks.py:260] loss = 1.122388, step = 26200 (3.096 sec)
I0804 19:40:02.208163 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0819
I0804 19:40:02.209790 140200711067520 basic_session_run_hooks.py:260] loss = 1.2204032, step = 26300 (3.117 sec)
I0804 19:40:05.263779 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7265
I0804 19:40:05.265557 140200711067520 basic_session_run_hooks.py:260] loss = 1.2088444, step = 26400 (3.056 sec)
I0804 19:40:08.357367 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3249
I0804 19:40:08.358909 140200711067520 basic_session_run_hooks.py:260] loss = 1.1157737, step = 26500 (3.093 sec)
I0804 19:40:11.458297 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2484
I0804 19:40:11.459793 140200711067520 basic_session_run_hooks.py:260] loss = 1.1500334, step = 26600 (3.101 sec)
I0804 19:40:14.575665 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0791
I0804 19:40:14.576990 140200711067520 basic_session_run_hooks.py:260] loss = 1.1873708, step = 26700 (3.117 sec)
I0804 19:40:17.696856 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0386
I0804 19:40:17.697938 140200711067520 basic_session_run_hooks.py:260] loss = 1.1324512, step = 26800 (3.121 sec)
I0804 19:40:20.828170 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9375
I0804 19:40:20.829654 140200711067520 basic_session_run_hooks.py:260] loss = 1.0946645, step = 26900 (3.132 sec)
I0804 19:40:23.903733 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 27000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:40:24.199718 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:40:24.240202 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.3061
I0804 19:40:24.241343 140200711067520 basic_session_run_hooks.py:260] loss = 1.1957022, step = 27000 (3.412 sec)
I0804 19:40:27.355068 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1045
I0804 19:40:27.356534 140200711067520 basic_session_run_hooks.py:260] loss = 1.0480574, step = 27100 (3.115 sec)
I0804 19:40:30.416153 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6679
I0804 19:40:30.417474 140200711067520 basic_session_run_hooks.py:260] loss = 1.2231941, step = 27200 (3.061 sec)
I0804 19:40:33.458863 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8656
I0804 19:40:33.460330 140200711067520 basic_session_run_hooks.py:260] loss = 1.1976117, step = 27300 (3.043 sec)
I0804 19:40:36.518341 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6852
I0804 19:40:36.520019 140200711067520 basic_session_run_hooks.py:260] loss = 1.0975667, step = 27400 (3.060 sec)
I0804 19:40:39.598225 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4688
I0804 19:40:39.599837 140200711067520 basic_session_run_hooks.py:260] loss = 1.0454144, step = 27500 (3.080 sec)
I0804 19:40:42.659654 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6644
I0804 19:40:42.661021 140200711067520 basic_session_run_hooks.py:260] loss = 1.1412435, step = 27600 (3.061 sec)
I0804 19:40:45.733619 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5314
I0804 19:40:45.735571 140200711067520 basic_session_run_hooks.py:260] loss = 1.1361846, step = 27700 (3.075 sec)
I0804 19:40:48.811731 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4874
I0804 19:40:48.813244 140200711067520 basic_session_run_hooks.py:260] loss = 1.1252506, step = 27800 (3.078 sec)
I0804 19:40:51.903401 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3449
I0804 19:40:51.905001 140200711067520 basic_session_run_hooks.py:260] loss = 1.2447684, step = 27900 (3.092 sec)
I0804 19:40:54.919029 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 28000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:40:55.202318 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:40:55.238695 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.9821
I0804 19:40:55.239888 140200711067520 basic_session_run_hooks.py:260] loss = 1.1910185, step = 28000 (3.335 sec)
I0804 19:40:58.289263 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.781
I0804 19:40:58.290777 140200711067520 basic_session_run_hooks.py:260] loss = 1.1956398, step = 28100 (3.051 sec)
I0804 19:41:01.331097 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.875
I0804 19:41:01.332472 140200711067520 basic_session_run_hooks.py:260] loss = 1.0986398, step = 28200 (3.042 sec)
I0804 19:41:04.409526 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4843
I0804 19:41:04.410885 140200711067520 basic_session_run_hooks.py:260] loss = 1.1499906, step = 28300 (3.078 sec)
I0804 19:41:07.469053 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6846
I0804 19:41:07.470686 140200711067520 basic_session_run_hooks.py:260] loss = 1.1346446, step = 28400 (3.060 sec)
I0804 19:41:10.508963 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8957
I0804 19:41:10.510234 140200711067520 basic_session_run_hooks.py:260] loss = 1.1012393, step = 28500 (3.040 sec)
I0804 19:41:13.556344 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.8152
I0804 19:41:13.557551 140200711067520 basic_session_run_hooks.py:260] loss = 1.1290898, step = 28600 (3.047 sec)
I0804 19:41:16.640304 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.426
I0804 19:41:16.641870 140200711067520 basic_session_run_hooks.py:260] loss = 1.0827153, step = 28700 (3.084 sec)
I0804 19:41:19.798603 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6626
I0804 19:41:19.799752 140200711067520 basic_session_run_hooks.py:260] loss = 1.2034243, step = 28800 (3.158 sec)
I0804 19:41:22.937272 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8606
I0804 19:41:22.938701 140200711067520 basic_session_run_hooks.py:260] loss = 1.1859727, step = 28900 (3.139 sec)
I0804 19:41:26.041780 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 29000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:41:26.325041 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:41:26.366203 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.1635
I0804 19:41:26.367583 140200711067520 basic_session_run_hooks.py:260] loss = 1.1626228, step = 29000 (3.429 sec)
I0804 19:41:29.550554 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4037
I0804 19:41:29.551915 140200711067520 basic_session_run_hooks.py:260] loss = 1.1033005, step = 29100 (3.184 sec)
I0804 19:41:32.649246 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2715
I0804 19:41:32.650592 140200711067520 basic_session_run_hooks.py:260] loss = 1.0244342, step = 29200 (3.099 sec)
I0804 19:41:35.724768 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5149
I0804 19:41:35.726013 140200711067520 basic_session_run_hooks.py:260] loss = 1.1233537, step = 29300 (3.075 sec)
I0804 19:41:38.855039 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9461
I0804 19:41:38.856492 140200711067520 basic_session_run_hooks.py:260] loss = 1.1355093, step = 29400 (3.130 sec)
I0804 19:41:41.988116 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9176
I0804 19:41:41.989664 140200711067520 basic_session_run_hooks.py:260] loss = 1.1635525, step = 29500 (3.133 sec)
I0804 19:41:45.132560 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8022
I0804 19:41:45.133991 140200711067520 basic_session_run_hooks.py:260] loss = 1.1484761, step = 29600 (3.144 sec)
I0804 19:41:48.272463 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8482
I0804 19:41:48.273803 140200711067520 basic_session_run_hooks.py:260] loss = 1.1287494, step = 29700 (3.140 sec)
I0804 19:41:51.417184 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.799
I0804 19:41:51.418999 140200711067520 basic_session_run_hooks.py:260] loss = 1.2534288, step = 29800 (3.145 sec)
I0804 19:41:54.538943 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0332
I0804 19:41:54.540071 140200711067520 basic_session_run_hooks.py:260] loss = 1.1392552, step = 29900 (3.121 sec)
I0804 19:41:57.571832 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 30000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:41:57.864303 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:41:57.904655 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.7113
I0804 19:41:57.906008 140200711067520 basic_session_run_hooks.py:260] loss = 1.2355024, step = 30000 (3.366 sec)
I0804 19:42:00.968649 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.6374
I0804 19:42:00.969892 140200711067520 basic_session_run_hooks.py:260] loss = 1.1847318, step = 30100 (3.064 sec)
I0804 19:42:04.054337 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4077
I0804 19:42:04.055823 140200711067520 basic_session_run_hooks.py:260] loss = 1.2119321, step = 30200 (3.086 sec)
I0804 19:42:07.164514 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1527
I0804 19:42:07.165837 140200711067520 basic_session_run_hooks.py:260] loss = 1.1538389, step = 30300 (3.110 sec)
I0804 19:42:10.285569 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0404
I0804 19:42:10.286929 140200711067520 basic_session_run_hooks.py:260] loss = 1.1768385, step = 30400 (3.121 sec)
I0804 19:42:13.418894 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9149
I0804 19:42:13.420933 140200711067520 basic_session_run_hooks.py:260] loss = 1.1893463, step = 30500 (3.134 sec)
I0804 19:42:16.553886 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.898
I0804 19:42:16.555094 140200711067520 basic_session_run_hooks.py:260] loss = 1.228784, step = 30600 (3.134 sec)
I0804 19:42:19.660362 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1908
I0804 19:42:19.661844 140200711067520 basic_session_run_hooks.py:260] loss = 1.1657281, step = 30700 (3.107 sec)
I0804 19:42:22.716851 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.7174
I0804 19:42:22.718348 140200711067520 basic_session_run_hooks.py:260] loss = 1.0922602, step = 30800 (3.057 sec)
I0804 19:42:25.808853 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3414
I0804 19:42:25.810248 140200711067520 basic_session_run_hooks.py:260] loss = 1.0505708, step = 30900 (3.092 sec)
I0804 19:42:28.861890 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 31000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:42:29.144351 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:42:29.179987 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.6633
I0804 19:42:29.181079 140200711067520 basic_session_run_hooks.py:260] loss = 1.1450275, step = 31000 (3.371 sec)
I0804 19:42:32.264278 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.4226
I0804 19:42:32.265774 140200711067520 basic_session_run_hooks.py:260] loss = 1.2153137, step = 31100 (3.085 sec)
I0804 19:42:35.357343 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3305
I0804 19:42:35.358888 140200711067520 basic_session_run_hooks.py:260] loss = 1.0850574, step = 31200 (3.093 sec)
I0804 19:42:38.470263 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.124
I0804 19:42:38.471723 140200711067520 basic_session_run_hooks.py:260] loss = 1.1064955, step = 31300 (3.113 sec)
I0804 19:42:41.562155 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3429
I0804 19:42:41.563751 140200711067520 basic_session_run_hooks.py:260] loss = 1.1050696, step = 31400 (3.092 sec)
I0804 19:42:44.660566 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2746
I0804 19:42:44.661798 140200711067520 basic_session_run_hooks.py:260] loss = 1.1563405, step = 31500 (3.098 sec)
I0804 19:42:47.734446 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5323
I0804 19:42:47.735753 140200711067520 basic_session_run_hooks.py:260] loss = 1.0993637, step = 31600 (3.074 sec)
I0804 19:42:50.858160 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.013
I0804 19:42:50.859621 140200711067520 basic_session_run_hooks.py:260] loss = 1.1257668, step = 31700 (3.124 sec)
I0804 19:42:53.947921 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.365
I0804 19:42:53.949494 140200711067520 basic_session_run_hooks.py:260] loss = 1.1356064, step = 31800 (3.090 sec)
I0804 19:42:57.064991 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0814
I0804 19:42:57.066840 140200711067520 basic_session_run_hooks.py:260] loss = 1.2098117, step = 31900 (3.117 sec)
I0804 19:43:00.166999 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 32000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:43:00.463842 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:43:00.505716 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0634
I0804 19:43:00.506840 140200711067520 basic_session_run_hooks.py:260] loss = 1.0860581, step = 32000 (3.440 sec)
I0804 19:43:03.620659 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1036
I0804 19:43:03.622276 140200711067520 basic_session_run_hooks.py:260] loss = 1.1525004, step = 32100 (3.115 sec)
I0804 19:43:06.723973 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2237
I0804 19:43:06.725259 140200711067520 basic_session_run_hooks.py:260] loss = 1.2531799, step = 32200 (3.103 sec)
I0804 19:43:09.861029 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8769
I0804 19:43:09.862672 140200711067520 basic_session_run_hooks.py:260] loss = 1.0775325, step = 32300 (3.137 sec)
I0804 19:43:12.964781 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2191
I0804 19:43:12.966395 140200711067520 basic_session_run_hooks.py:260] loss = 1.1300572, step = 32400 (3.104 sec)
I0804 19:43:16.114838 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7463
I0804 19:43:16.116972 140200711067520 basic_session_run_hooks.py:260] loss = 1.1692545, step = 32500 (3.151 sec)
I0804 19:43:19.249599 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8998
I0804 19:43:19.251115 140200711067520 basic_session_run_hooks.py:260] loss = 1.1501764, step = 32600 (3.134 sec)
I0804 19:43:22.449326 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2524
I0804 19:43:22.450523 140200711067520 basic_session_run_hooks.py:260] loss = 1.0724894, step = 32700 (3.199 sec)
I0804 19:43:25.615726 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5817
I0804 19:43:25.617092 140200711067520 basic_session_run_hooks.py:260] loss = 1.1188151, step = 32800 (3.167 sec)
I0804 19:43:28.802237 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3822
I0804 19:43:28.803824 140200711067520 basic_session_run_hooks.py:260] loss = 1.0858948, step = 32900 (3.187 sec)
I0804 19:43:31.968640 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 33000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:43:32.254656 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:43:32.297674 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6085
I0804 19:43:32.298858 140200711067520 basic_session_run_hooks.py:260] loss = 1.1348358, step = 33000 (3.495 sec)
I0804 19:43:35.428545 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9404
I0804 19:43:35.430008 140200711067520 basic_session_run_hooks.py:260] loss = 1.1708028, step = 33100 (3.131 sec)
I0804 19:43:38.543473 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1033
I0804 19:43:38.544840 140200711067520 basic_session_run_hooks.py:260] loss = 1.1397612, step = 33200 (3.115 sec)
I0804 19:43:41.643160 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2611
I0804 19:43:41.644580 140200711067520 basic_session_run_hooks.py:260] loss = 1.1923155, step = 33300 (3.100 sec)
I0804 19:43:44.753402 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1519
I0804 19:43:44.755007 140200711067520 basic_session_run_hooks.py:260] loss = 1.1914612, step = 33400 (3.110 sec)
I0804 19:43:47.880235 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9813
I0804 19:43:47.881893 140200711067520 basic_session_run_hooks.py:260] loss = 1.1400195, step = 33500 (3.127 sec)
I0804 19:43:51.039673 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.651
I0804 19:43:51.041110 140200711067520 basic_session_run_hooks.py:260] loss = 1.0916404, step = 33600 (3.159 sec)
I0804 19:43:54.177462 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.87
I0804 19:43:54.178823 140200711067520 basic_session_run_hooks.py:260] loss = 1.174411, step = 33700 (3.138 sec)
I0804 19:43:57.316867 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8529
I0804 19:43:57.318130 140200711067520 basic_session_run_hooks.py:260] loss = 1.0537717, step = 33800 (3.139 sec)
I0804 19:44:00.442864 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9898
I0804 19:44:00.444636 140200711067520 basic_session_run_hooks.py:260] loss = 1.075204, step = 33900 (3.127 sec)
I0804 19:44:03.517956 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 34000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:44:03.807043 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:44:03.852844 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.3255
I0804 19:44:03.854137 140200711067520 basic_session_run_hooks.py:260] loss = 1.1966132, step = 34000 (3.410 sec)
I0804 19:44:06.974815 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0312
I0804 19:44:06.976236 140200711067520 basic_session_run_hooks.py:260] loss = 1.0749761, step = 34100 (3.122 sec)
I0804 19:44:10.098189 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0167
I0804 19:44:10.099615 140200711067520 basic_session_run_hooks.py:260] loss = 1.1096065, step = 34200 (3.123 sec)
I0804 19:44:13.216823 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0654
I0804 19:44:13.217978 140200711067520 basic_session_run_hooks.py:260] loss = 1.1340765, step = 34300 (3.118 sec)
I0804 19:44:16.311789 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3106
I0804 19:44:16.313415 140200711067520 basic_session_run_hooks.py:260] loss = 1.1393902, step = 34400 (3.095 sec)
I0804 19:44:19.422606 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.146
I0804 19:44:19.423888 140200711067520 basic_session_run_hooks.py:260] loss = 1.085156, step = 34500 (3.110 sec)
I0804 19:44:22.530996 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1708
I0804 19:44:22.532528 140200711067520 basic_session_run_hooks.py:260] loss = 1.1606448, step = 34600 (3.109 sec)
I0804 19:44:25.656658 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.993
I0804 19:44:25.658032 140200711067520 basic_session_run_hooks.py:260] loss = 1.1447561, step = 34700 (3.126 sec)
I0804 19:44:28.768217 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1382
I0804 19:44:28.769623 140200711067520 basic_session_run_hooks.py:260] loss = 1.0503354, step = 34800 (3.112 sec)
I0804 19:44:31.890390 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0291
I0804 19:44:31.891864 140200711067520 basic_session_run_hooks.py:260] loss = 1.1648916, step = 34900 (3.122 sec)
I0804 19:44:34.982296 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 35000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:44:35.262665 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:44:35.304812 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.2873
I0804 19:44:35.305959 140200711067520 basic_session_run_hooks.py:260] loss = 1.0839369, step = 35000 (3.414 sec)
I0804 19:44:38.438213 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9145
I0804 19:44:38.439763 140200711067520 basic_session_run_hooks.py:260] loss = 1.1892983, step = 35100 (3.134 sec)
I0804 19:44:41.581030 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8186
I0804 19:44:41.582527 140200711067520 basic_session_run_hooks.py:260] loss = 1.1458148, step = 35200 (3.143 sec)
I0804 19:44:44.717597 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8824
I0804 19:44:44.718990 140200711067520 basic_session_run_hooks.py:260] loss = 1.0933186, step = 35300 (3.136 sec)
I0804 19:44:47.861055 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8118
I0804 19:44:47.862513 140200711067520 basic_session_run_hooks.py:260] loss = 1.0847974, step = 35400 (3.144 sec)
I0804 19:44:50.949373 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.38
I0804 19:44:50.950903 140200711067520 basic_session_run_hooks.py:260] loss = 1.1280469, step = 35500 (3.088 sec)
I0804 19:44:54.023162 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5332
I0804 19:44:54.024474 140200711067520 basic_session_run_hooks.py:260] loss = 1.1532847, step = 35600 (3.074 sec)
I0804 19:44:57.111187 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3831
I0804 19:44:57.112462 140200711067520 basic_session_run_hooks.py:260] loss = 1.1384014, step = 35700 (3.088 sec)
I0804 19:45:00.198591 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3898
I0804 19:45:00.199975 140200711067520 basic_session_run_hooks.py:260] loss = 1.1691403, step = 35800 (3.088 sec)
I0804 19:45:03.296027 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2847
I0804 19:45:03.297636 140200711067520 basic_session_run_hooks.py:260] loss = 1.1008931, step = 35900 (3.098 sec)
I0804 19:45:06.357104 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 36000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:45:06.634567 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:45:06.682233 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.5316
I0804 19:45:06.683579 140200711067520 basic_session_run_hooks.py:260] loss = 1.117098, step = 36000 (3.386 sec)
I0804 19:45:09.769646 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3895
I0804 19:45:09.770912 140200711067520 basic_session_run_hooks.py:260] loss = 1.1830325, step = 36100 (3.087 sec)
I0804 19:45:12.885581 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0933
I0804 19:45:12.886886 140200711067520 basic_session_run_hooks.py:260] loss = 1.0935432, step = 36200 (3.116 sec)
I0804 19:45:15.983802 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2765
I0804 19:45:15.984927 140200711067520 basic_session_run_hooks.py:260] loss = 1.0860431, step = 36300 (3.098 sec)
I0804 19:45:19.104549 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0437
I0804 19:45:19.105807 140200711067520 basic_session_run_hooks.py:260] loss = 1.1255058, step = 36400 (3.121 sec)
I0804 19:45:22.252201 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7697
I0804 19:45:22.253720 140200711067520 basic_session_run_hooks.py:260] loss = 1.2379012, step = 36500 (3.148 sec)
I0804 19:45:25.402438 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7437
I0804 19:45:25.404094 140200711067520 basic_session_run_hooks.py:260] loss = 1.0574921, step = 36600 (3.150 sec)
I0804 19:45:28.547827 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7924
I0804 19:45:28.549195 140200711067520 basic_session_run_hooks.py:260] loss = 1.1052928, step = 36700 (3.145 sec)
I0804 19:45:31.687101 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8545
I0804 19:45:31.688293 140200711067520 basic_session_run_hooks.py:260] loss = 1.1577585, step = 36800 (3.139 sec)
I0804 19:45:34.848098 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6356
I0804 19:45:34.849531 140200711067520 basic_session_run_hooks.py:260] loss = 1.1120983, step = 36900 (3.161 sec)
I0804 19:45:37.946725 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 37000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:45:38.235222 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 19:45:38.236586 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 19:45:38.382125 140200711067520 estimator.py:1145] Calling model_fn.
I0804 19:45:38.383153 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 19:45:38.383559 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 19:45:38.383653 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 19:45:38.383737 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 19:45:38.383805 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 19:45:38.383888 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 19:45:38.383961 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 19:45:38.472891 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 19:45:38.533082 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 19:45:38.933140 140200711067520 t2t_model.py:2172] Building model body
I0804 19:45:39.630722 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 19:45:40.347840 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 19:45:40.365670 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T19:45:40Z
I0804 19:45:40.738870 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 19:45:40.739563: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:45:40.739961: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 19:45:40.740060: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 19:45:40.740084: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 19:45:40.740105: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 19:45:40.740129: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 19:45:40.740150: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 19:45:40.740168: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 19:45:40.740189: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 19:45:40.740296: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:45:40.740756: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:45:40.741071: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 19:45:40.741113: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 19:45:40.741126: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 19:45:40.741137: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 19:45:40.741432: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:45:40.741829: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:45:40.742154: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 19:45:40.743547 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-37000
I0804 19:45:40.943182 140200711067520 session_manager.py:500] Running local_init_op.
I0804 19:45:40.999976 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 19:45:46.976284 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 19:45:52.339315 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 19:45:57.674107 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 19:46:03.029103 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 19:46:08.373158 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 19:46:13.735360 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 19:46:19.109217 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 19:46:24.490439 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 19:46:29.859190 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 19:46:34.737031 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-19:46:34
I0804 19:46:34.737284 140200711067520 estimator.py:2039] Saving dict for global step 37000: global_step = 37000, loss = 1.229527, metrics-paper_generation_problem/targets/accuracy = 0.659651, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8769182, metrics-paper_generation_problem/targets/approx_bleu_score = 0.46940625, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.2295651, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.56771475, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.68329483
I0804 19:46:34.737883 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 37000: experiment/transformer/transformer_small/output/model.ckpt-37000
I0804 19:46:34.792580 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.66821
I0804 19:46:34.793887 140200711067520 basic_session_run_hooks.py:260] loss = 1.0569671, step = 37000 (59.944 sec)
I0804 19:46:38.017661 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0073
I0804 19:46:38.018877 140200711067520 basic_session_run_hooks.py:260] loss = 1.0760849, step = 37100 (3.225 sec)
I0804 19:46:41.173192 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6904
I0804 19:46:41.174814 140200711067520 basic_session_run_hooks.py:260] loss = 1.1021217, step = 37200 (3.156 sec)
I0804 19:46:44.356107 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4179
I0804 19:46:44.357934 140200711067520 basic_session_run_hooks.py:260] loss = 1.0698866, step = 37300 (3.183 sec)
I0804 19:46:47.529648 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5104
I0804 19:46:47.530972 140200711067520 basic_session_run_hooks.py:260] loss = 1.0578548, step = 37400 (3.173 sec)
I0804 19:46:50.712347 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4199
I0804 19:46:50.713885 140200711067520 basic_session_run_hooks.py:260] loss = 1.1988635, step = 37500 (3.183 sec)
I0804 19:46:53.902745 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3441
I0804 19:46:53.904009 140200711067520 basic_session_run_hooks.py:260] loss = 1.080587, step = 37600 (3.190 sec)
I0804 19:46:57.119250 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0897
I0804 19:46:57.120725 140200711067520 basic_session_run_hooks.py:260] loss = 1.1148747, step = 37700 (3.217 sec)
I0804 19:47:00.303402 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4054
I0804 19:47:00.304581 140200711067520 basic_session_run_hooks.py:260] loss = 1.1601979, step = 37800 (3.184 sec)
I0804 19:47:03.477780 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5022
I0804 19:47:03.479143 140200711067520 basic_session_run_hooks.py:260] loss = 1.1374818, step = 37900 (3.175 sec)
I0804 19:47:06.595283 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 38000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:47:06.905234 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:47:06.942517 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8624
I0804 19:47:06.943650 140200711067520 basic_session_run_hooks.py:260] loss = 1.1260904, step = 38000 (3.465 sec)
I0804 19:47:10.088244 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7889
I0804 19:47:10.089735 140200711067520 basic_session_run_hooks.py:260] loss = 1.1389525, step = 38100 (3.146 sec)
I0804 19:47:13.223808 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8923
I0804 19:47:13.225242 140200711067520 basic_session_run_hooks.py:260] loss = 1.1431541, step = 38200 (3.136 sec)
I0804 19:47:16.329014 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.204
I0804 19:47:16.330395 140200711067520 basic_session_run_hooks.py:260] loss = 1.1315507, step = 38300 (3.105 sec)
I0804 19:47:19.493152 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6042
I0804 19:47:19.494551 140200711067520 basic_session_run_hooks.py:260] loss = 1.165114, step = 38400 (3.164 sec)
I0804 19:47:22.641691 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7607
I0804 19:47:22.643098 140200711067520 basic_session_run_hooks.py:260] loss = 1.1816443, step = 38500 (3.149 sec)
I0804 19:47:25.753604 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1351
I0804 19:47:25.755117 140200711067520 basic_session_run_hooks.py:260] loss = 1.1648796, step = 38600 (3.112 sec)
I0804 19:47:28.889210 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8914
I0804 19:47:28.890955 140200711067520 basic_session_run_hooks.py:260] loss = 1.2497754, step = 38700 (3.136 sec)
I0804 19:47:32.012689 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0154
I0804 19:47:32.014197 140200711067520 basic_session_run_hooks.py:260] loss = 1.0903958, step = 38800 (3.123 sec)
I0804 19:47:35.111797 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2676
I0804 19:47:35.113016 140200711067520 basic_session_run_hooks.py:260] loss = 1.1297072, step = 38900 (3.099 sec)
I0804 19:47:38.259908 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 39000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:47:38.552392 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:47:38.594200 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7155
I0804 19:47:38.595472 140200711067520 basic_session_run_hooks.py:260] loss = 1.1950412, step = 39000 (3.482 sec)
I0804 19:47:41.670649 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5053
I0804 19:47:41.671877 140200711067520 basic_session_run_hooks.py:260] loss = 1.0435718, step = 39100 (3.076 sec)
I0804 19:47:44.773923 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2241
I0804 19:47:44.775115 140200711067520 basic_session_run_hooks.py:260] loss = 1.1355599, step = 39200 (3.103 sec)
I0804 19:47:47.886404 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1287
I0804 19:47:47.887532 140200711067520 basic_session_run_hooks.py:260] loss = 1.1106049, step = 39300 (3.112 sec)
I0804 19:47:51.002951 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0868
I0804 19:47:51.004284 140200711067520 basic_session_run_hooks.py:260] loss = 1.162496, step = 39400 (3.117 sec)
I0804 19:47:54.122167 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0594
I0804 19:47:54.123507 140200711067520 basic_session_run_hooks.py:260] loss = 1.1844639, step = 39500 (3.119 sec)
I0804 19:47:57.259527 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.874
I0804 19:47:57.260790 140200711067520 basic_session_run_hooks.py:260] loss = 1.117798, step = 39600 (3.137 sec)
I0804 19:48:00.427068 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5701
I0804 19:48:00.428402 140200711067520 basic_session_run_hooks.py:260] loss = 1.0189216, step = 39700 (3.168 sec)
I0804 19:48:03.608251 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4347
I0804 19:48:03.609885 140200711067520 basic_session_run_hooks.py:260] loss = 1.0835203, step = 39800 (3.181 sec)
I0804 19:48:06.767454 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6539
I0804 19:48:06.768878 140200711067520 basic_session_run_hooks.py:260] loss = 1.1544534, step = 39900 (3.159 sec)
I0804 19:48:09.943732 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 40000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:48:10.243828 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:48:10.291788 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.374
I0804 19:48:10.293133 140200711067520 basic_session_run_hooks.py:260] loss = 1.0612311, step = 40000 (3.524 sec)
I0804 19:48:13.465312 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5108
I0804 19:48:13.466769 140200711067520 basic_session_run_hooks.py:260] loss = 1.1926965, step = 40100 (3.174 sec)
I0804 19:48:16.613603 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7633
I0804 19:48:16.614854 140200711067520 basic_session_run_hooks.py:260] loss = 1.101234, step = 40200 (3.148 sec)
I0804 19:48:19.799082 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3925
I0804 19:48:19.800540 140200711067520 basic_session_run_hooks.py:260] loss = 1.0723251, step = 40300 (3.186 sec)
I0804 19:48:22.979838 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4388
I0804 19:48:22.981451 140200711067520 basic_session_run_hooks.py:260] loss = 1.1735097, step = 40400 (3.181 sec)
I0804 19:48:26.169119 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3552
I0804 19:48:26.170618 140200711067520 basic_session_run_hooks.py:260] loss = 1.10444, step = 40500 (3.189 sec)
I0804 19:48:29.295563 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9853
I0804 19:48:29.296746 140200711067520 basic_session_run_hooks.py:260] loss = 1.1231445, step = 40600 (3.126 sec)
I0804 19:48:32.397540 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2375
I0804 19:48:32.399085 140200711067520 basic_session_run_hooks.py:260] loss = 1.0709445, step = 40700 (3.102 sec)
I0804 19:48:35.526154 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9629
I0804 19:48:35.527490 140200711067520 basic_session_run_hooks.py:260] loss = 1.1798171, step = 40800 (3.128 sec)
I0804 19:48:38.651607 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9954
I0804 19:48:38.652792 140200711067520 basic_session_run_hooks.py:260] loss = 1.1527029, step = 40900 (3.125 sec)
I0804 19:48:41.772135 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 41000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:48:42.078557 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:48:42.118563 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8435
I0804 19:48:42.119494 140200711067520 basic_session_run_hooks.py:260] loss = 1.0874952, step = 41000 (3.467 sec)
I0804 19:48:45.211766 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3291
I0804 19:48:45.213292 140200711067520 basic_session_run_hooks.py:260] loss = 1.0950297, step = 41100 (3.094 sec)
I0804 19:48:48.313573 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2396
I0804 19:48:48.315020 140200711067520 basic_session_run_hooks.py:260] loss = 1.0911891, step = 41200 (3.102 sec)
I0804 19:48:51.444739 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9367
I0804 19:48:51.446208 140200711067520 basic_session_run_hooks.py:260] loss = 1.1207026, step = 41300 (3.131 sec)
I0804 19:48:54.554536 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1566
I0804 19:48:54.556198 140200711067520 basic_session_run_hooks.py:260] loss = 1.156787, step = 41400 (3.110 sec)
I0804 19:48:57.679709 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9984
I0804 19:48:57.681115 140200711067520 basic_session_run_hooks.py:260] loss = 1.1182648, step = 41500 (3.125 sec)
I0804 19:49:00.818192 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8623
I0804 19:49:00.819619 140200711067520 basic_session_run_hooks.py:260] loss = 1.0795518, step = 41600 (3.139 sec)
I0804 19:49:03.937254 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0608
I0804 19:49:03.938905 140200711067520 basic_session_run_hooks.py:260] loss = 1.14291, step = 41700 (3.119 sec)
I0804 19:49:07.102210 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.596
I0804 19:49:07.103646 140200711067520 basic_session_run_hooks.py:260] loss = 1.1765906, step = 41800 (3.165 sec)
I0804 19:49:10.240308 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8664
I0804 19:49:10.241769 140200711067520 basic_session_run_hooks.py:260] loss = 1.1178077, step = 41900 (3.138 sec)
I0804 19:49:13.346736 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 42000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:49:13.636853 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:49:13.692546 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.9669
I0804 19:49:13.694123 140200711067520 basic_session_run_hooks.py:260] loss = 1.0861866, step = 42000 (3.452 sec)
I0804 19:49:16.857036 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6004
I0804 19:49:16.858393 140200711067520 basic_session_run_hooks.py:260] loss = 1.1167343, step = 42100 (3.164 sec)
I0804 19:49:20.019611 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6202
I0804 19:49:20.020806 140200711067520 basic_session_run_hooks.py:260] loss = 1.0742412, step = 42200 (3.162 sec)
I0804 19:49:23.205010 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3928
I0804 19:49:23.206369 140200711067520 basic_session_run_hooks.py:260] loss = 1.2063092, step = 42300 (3.186 sec)
I0804 19:49:26.357926 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7166
I0804 19:49:26.359014 140200711067520 basic_session_run_hooks.py:260] loss = 1.0842273, step = 42400 (3.153 sec)
I0804 19:49:29.524014 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5849
I0804 19:49:29.525551 140200711067520 basic_session_run_hooks.py:260] loss = 1.1359447, step = 42500 (3.167 sec)
I0804 19:49:32.696633 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5199
I0804 19:49:32.698197 140200711067520 basic_session_run_hooks.py:260] loss = 1.1206957, step = 42600 (3.173 sec)
I0804 19:49:35.863847 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5732
I0804 19:49:35.865286 140200711067520 basic_session_run_hooks.py:260] loss = 1.1611115, step = 42700 (3.167 sec)
I0804 19:49:39.041852 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4664
I0804 19:49:39.043067 140200711067520 basic_session_run_hooks.py:260] loss = 1.1207973, step = 42800 (3.178 sec)
I0804 19:49:42.214304 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5213
I0804 19:49:42.215958 140200711067520 basic_session_run_hooks.py:260] loss = 1.0937647, step = 42900 (3.173 sec)
I0804 19:49:45.310682 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 43000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:49:45.614023 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:49:45.654407 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0686
I0804 19:49:45.655745 140200711067520 basic_session_run_hooks.py:260] loss = 1.134537, step = 43000 (3.440 sec)
I0804 19:49:48.774170 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0541
I0804 19:49:48.775308 140200711067520 basic_session_run_hooks.py:260] loss = 1.053636, step = 43100 (3.120 sec)
I0804 19:49:51.903896 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9516
I0804 19:49:51.905274 140200711067520 basic_session_run_hooks.py:260] loss = 1.1231464, step = 43200 (3.130 sec)
I0804 19:49:55.016341 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1292
I0804 19:49:55.017774 140200711067520 basic_session_run_hooks.py:260] loss = 1.1810625, step = 43300 (3.112 sec)
I0804 19:49:58.130929 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.107
I0804 19:49:58.132284 140200711067520 basic_session_run_hooks.py:260] loss = 1.1416055, step = 43400 (3.115 sec)
I0804 19:50:01.255375 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0057
I0804 19:50:01.257084 140200711067520 basic_session_run_hooks.py:260] loss = 1.0467563, step = 43500 (3.125 sec)
I0804 19:50:04.385756 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.945
I0804 19:50:04.387395 140200711067520 basic_session_run_hooks.py:260] loss = 1.1801257, step = 43600 (3.130 sec)
I0804 19:50:07.543190 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6712
I0804 19:50:07.544659 140200711067520 basic_session_run_hooks.py:260] loss = 1.0968823, step = 43700 (3.157 sec)
I0804 19:50:10.706562 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.612
I0804 19:50:10.707848 140200711067520 basic_session_run_hooks.py:260] loss = 1.0981392, step = 43800 (3.163 sec)
I0804 19:50:13.839834 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9154
I0804 19:50:13.841026 140200711067520 basic_session_run_hooks.py:260] loss = 1.2310303, step = 43900 (3.133 sec)
I0804 19:50:16.967529 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 44000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:50:17.272054 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:50:17.312608 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7951
I0804 19:50:17.313704 140200711067520 basic_session_run_hooks.py:260] loss = 1.0916755, step = 44000 (3.473 sec)
I0804 19:50:20.499437 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3798
I0804 19:50:20.500951 140200711067520 basic_session_run_hooks.py:260] loss = 1.0956663, step = 44100 (3.187 sec)
I0804 19:50:23.657495 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6647
I0804 19:50:23.658710 140200711067520 basic_session_run_hooks.py:260] loss = 1.1110739, step = 44200 (3.158 sec)
I0804 19:50:26.822151 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5991
I0804 19:50:26.823799 140200711067520 basic_session_run_hooks.py:260] loss = 1.1317707, step = 44300 (3.165 sec)
I0804 19:50:29.998598 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4816
I0804 19:50:30.000324 140200711067520 basic_session_run_hooks.py:260] loss = 1.1228254, step = 44400 (3.177 sec)
I0804 19:50:33.181403 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4189
I0804 19:50:33.183091 140200711067520 basic_session_run_hooks.py:260] loss = 1.0867153, step = 44500 (3.183 sec)
I0804 19:50:36.288378 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1856
I0804 19:50:36.290023 140200711067520 basic_session_run_hooks.py:260] loss = 1.127906, step = 44600 (3.107 sec)
I0804 19:50:39.455099 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5785
I0804 19:50:39.456608 140200711067520 basic_session_run_hooks.py:260] loss = 1.1027145, step = 44700 (3.167 sec)
I0804 19:50:42.621823 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5782
I0804 19:50:42.623377 140200711067520 basic_session_run_hooks.py:260] loss = 1.0924087, step = 44800 (3.167 sec)
I0804 19:50:45.814135 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3251
I0804 19:50:45.815702 140200711067520 basic_session_run_hooks.py:260] loss = 1.0879875, step = 44900 (3.192 sec)
I0804 19:50:48.975697 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 45000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:50:49.277351 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:50:49.313450 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.577
I0804 19:50:49.314517 140200711067520 basic_session_run_hooks.py:260] loss = 1.03856, step = 45000 (3.499 sec)
I0804 19:50:52.505134 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3314
I0804 19:50:52.506494 140200711067520 basic_session_run_hooks.py:260] loss = 1.1044792, step = 45100 (3.192 sec)
I0804 19:50:55.674640 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5507
I0804 19:50:55.675972 140200711067520 basic_session_run_hooks.py:260] loss = 1.1221156, step = 45200 (3.169 sec)
I0804 19:50:58.880743 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1905
I0804 19:50:58.882122 140200711067520 basic_session_run_hooks.py:260] loss = 1.1480699, step = 45300 (3.206 sec)
I0804 19:51:02.014692 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9087
I0804 19:51:02.016048 140200711067520 basic_session_run_hooks.py:260] loss = 1.2268764, step = 45400 (3.134 sec)
I0804 19:51:05.178883 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6036
I0804 19:51:05.180268 140200711067520 basic_session_run_hooks.py:260] loss = 1.0690833, step = 45500 (3.164 sec)
I0804 19:51:08.329596 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7391
I0804 19:51:08.331135 140200711067520 basic_session_run_hooks.py:260] loss = 1.1909672, step = 45600 (3.151 sec)
I0804 19:51:11.484272 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6988
I0804 19:51:11.485639 140200711067520 basic_session_run_hooks.py:260] loss = 1.1587971, step = 45700 (3.155 sec)
I0804 19:51:14.633188 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7569
I0804 19:51:14.634793 140200711067520 basic_session_run_hooks.py:260] loss = 1.125182, step = 45800 (3.149 sec)
I0804 19:51:17.791949 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.658
I0804 19:51:17.793368 140200711067520 basic_session_run_hooks.py:260] loss = 1.0950791, step = 45900 (3.159 sec)
I0804 19:51:20.893923 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 46000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:51:21.196321 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:51:21.235576 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0393
I0804 19:51:21.236660 140200711067520 basic_session_run_hooks.py:260] loss = 1.1432881, step = 46000 (3.443 sec)
I0804 19:51:24.413073 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.471
I0804 19:51:24.414394 140200711067520 basic_session_run_hooks.py:260] loss = 1.147024, step = 46100 (3.178 sec)
I0804 19:51:27.545258 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9265
I0804 19:51:27.546638 140200711067520 basic_session_run_hooks.py:260] loss = 1.0977383, step = 46200 (3.132 sec)
I0804 19:51:30.701796 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6805
I0804 19:51:30.703228 140200711067520 basic_session_run_hooks.py:260] loss = 1.1768712, step = 46300 (3.157 sec)
I0804 19:51:33.847860 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7856
I0804 19:51:33.849293 140200711067520 basic_session_run_hooks.py:260] loss = 1.0523359, step = 46400 (3.146 sec)
I0804 19:51:37.027134 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4538
I0804 19:51:37.028738 140200711067520 basic_session_run_hooks.py:260] loss = 1.1298066, step = 46500 (3.179 sec)
I0804 19:51:40.198621 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5309
I0804 19:51:40.199979 140200711067520 basic_session_run_hooks.py:260] loss = 1.1377465, step = 46600 (3.171 sec)
I0804 19:51:43.375727 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4752
I0804 19:51:43.377113 140200711067520 basic_session_run_hooks.py:260] loss = 1.032058, step = 46700 (3.177 sec)
I0804 19:51:46.553083 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4725
I0804 19:51:46.554606 140200711067520 basic_session_run_hooks.py:260] loss = 1.1122661, step = 46800 (3.177 sec)
I0804 19:51:49.726198 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5149
I0804 19:51:49.727697 140200711067520 basic_session_run_hooks.py:260] loss = 1.1015502, step = 46900 (3.173 sec)
I0804 19:51:52.901107 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 47000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:51:53.200304 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:51:53.239494 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4633
I0804 19:51:53.240664 140200711067520 basic_session_run_hooks.py:260] loss = 1.157657, step = 47000 (3.513 sec)
I0804 19:51:56.423747 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4046
I0804 19:51:56.425590 140200711067520 basic_session_run_hooks.py:260] loss = 1.1277972, step = 47100 (3.185 sec)
I0804 19:51:59.626657 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2215
I0804 19:51:59.628106 140200711067520 basic_session_run_hooks.py:260] loss = 1.1428422, step = 47200 (3.203 sec)
I0804 19:52:02.855495 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9711
I0804 19:52:02.856688 140200711067520 basic_session_run_hooks.py:260] loss = 1.1522713, step = 47300 (3.229 sec)
I0804 19:52:06.074794 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0625
I0804 19:52:06.076230 140200711067520 basic_session_run_hooks.py:260] loss = 1.2067418, step = 47400 (3.220 sec)
I0804 19:52:09.317771 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.836
I0804 19:52:09.319367 140200711067520 basic_session_run_hooks.py:260] loss = 1.1119735, step = 47500 (3.243 sec)
I0804 19:52:12.552078 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9185
I0804 19:52:12.553522 140200711067520 basic_session_run_hooks.py:260] loss = 1.1338612, step = 47600 (3.234 sec)
I0804 19:52:15.699871 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7683
I0804 19:52:15.701689 140200711067520 basic_session_run_hooks.py:260] loss = 1.1050823, step = 47700 (3.148 sec)
I0804 19:52:18.794758 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3113
I0804 19:52:18.796313 140200711067520 basic_session_run_hooks.py:260] loss = 1.0467414, step = 47800 (3.095 sec)
I0804 19:52:21.963702 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5562
I0804 19:52:21.964854 140200711067520 basic_session_run_hooks.py:260] loss = 1.1535015, step = 47900 (3.169 sec)
I0804 19:52:25.042914 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 48000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:52:25.350454 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:52:25.388362 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.1998
I0804 19:52:25.389604 140200711067520 basic_session_run_hooks.py:260] loss = 1.0421687, step = 48000 (3.425 sec)
I0804 19:52:28.550123 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6283
I0804 19:52:28.551703 140200711067520 basic_session_run_hooks.py:260] loss = 1.1511866, step = 48100 (3.162 sec)
I0804 19:52:31.705336 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6936
I0804 19:52:31.706821 140200711067520 basic_session_run_hooks.py:260] loss = 1.0501173, step = 48200 (3.155 sec)
I0804 19:52:34.866969 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6292
I0804 19:52:34.868681 140200711067520 basic_session_run_hooks.py:260] loss = 1.0893638, step = 48300 (3.162 sec)
I0804 19:52:38.030803 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.607
I0804 19:52:38.032208 140200711067520 basic_session_run_hooks.py:260] loss = 1.141914, step = 48400 (3.164 sec)
I0804 19:52:41.162038 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9365
I0804 19:52:41.163757 140200711067520 basic_session_run_hooks.py:260] loss = 1.0813749, step = 48500 (3.132 sec)
I0804 19:52:44.272119 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1535
I0804 19:52:44.273633 140200711067520 basic_session_run_hooks.py:260] loss = 1.0449059, step = 48600 (3.110 sec)
I0804 19:52:47.428285 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.687
I0804 19:52:47.429540 140200711067520 basic_session_run_hooks.py:260] loss = 1.1289667, step = 48700 (3.156 sec)
I0804 19:52:50.605352 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4727
I0804 19:52:50.606736 140200711067520 basic_session_run_hooks.py:260] loss = 1.1132381, step = 48800 (3.177 sec)
I0804 19:52:53.817821 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1286
I0804 19:52:53.819218 140200711067520 basic_session_run_hooks.py:260] loss = 1.180058, step = 48900 (3.213 sec)
I0804 19:52:56.948262 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 49000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:52:57.242099 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:52:57.278123 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8991
I0804 19:52:57.279270 140200711067520 basic_session_run_hooks.py:260] loss = 1.1948634, step = 49000 (3.460 sec)
I0804 19:53:00.407149 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9591
I0804 19:53:00.408528 140200711067520 basic_session_run_hooks.py:260] loss = 1.1193062, step = 49100 (3.129 sec)
I0804 19:53:03.540292 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9168
I0804 19:53:03.541974 140200711067520 basic_session_run_hooks.py:260] loss = 1.103634, step = 49200 (3.133 sec)
I0804 19:53:06.674734 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9065
I0804 19:53:06.675946 140200711067520 basic_session_run_hooks.py:260] loss = 1.0827019, step = 49300 (3.134 sec)
I0804 19:53:09.828164 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7087
I0804 19:53:09.829476 140200711067520 basic_session_run_hooks.py:260] loss = 1.1328102, step = 49400 (3.154 sec)
I0804 19:53:12.983190 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6955
I0804 19:53:12.984576 140200711067520 basic_session_run_hooks.py:260] loss = 1.0673382, step = 49500 (3.155 sec)
I0804 19:53:16.118303 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8968
I0804 19:53:16.120167 140200711067520 basic_session_run_hooks.py:260] loss = 1.1392004, step = 49600 (3.136 sec)
I0804 19:53:19.292944 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4996
I0804 19:53:19.294044 140200711067520 basic_session_run_hooks.py:260] loss = 1.143987, step = 49700 (3.174 sec)
I0804 19:53:22.479009 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3867
I0804 19:53:22.480456 140200711067520 basic_session_run_hooks.py:260] loss = 1.1458966, step = 49800 (3.186 sec)
I0804 19:53:25.652931 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5067
I0804 19:53:25.654374 140200711067520 basic_session_run_hooks.py:260] loss = 1.1046932, step = 49900 (3.174 sec)
I0804 19:53:28.778250 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 50000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:53:29.071360 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:53:29.111127 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.9165
I0804 19:53:29.112239 140200711067520 basic_session_run_hooks.py:260] loss = 1.0981588, step = 50000 (3.458 sec)
I0804 19:53:32.281390 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5434
I0804 19:53:32.282866 140200711067520 basic_session_run_hooks.py:260] loss = 1.0612161, step = 50100 (3.171 sec)
I0804 19:53:35.403103 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0337
I0804 19:53:35.404511 140200711067520 basic_session_run_hooks.py:260] loss = 1.0447237, step = 50200 (3.122 sec)
I0804 19:53:38.542330 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8549
I0804 19:53:38.543864 140200711067520 basic_session_run_hooks.py:260] loss = 1.1472294, step = 50300 (3.139 sec)
I0804 19:53:41.687063 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7992
I0804 19:53:41.688698 140200711067520 basic_session_run_hooks.py:260] loss = 1.0845361, step = 50400 (3.145 sec)
I0804 19:53:44.903306 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0923
I0804 19:53:44.904629 140200711067520 basic_session_run_hooks.py:260] loss = 1.0767462, step = 50500 (3.216 sec)
I0804 19:53:48.074398 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5347
I0804 19:53:48.075791 140200711067520 basic_session_run_hooks.py:260] loss = 1.1957313, step = 50600 (3.171 sec)
I0804 19:53:51.256019 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4307
I0804 19:53:51.257282 140200711067520 basic_session_run_hooks.py:260] loss = 1.1528841, step = 50700 (3.181 sec)
I0804 19:53:54.428245 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5237
I0804 19:53:54.429663 140200711067520 basic_session_run_hooks.py:260] loss = 1.1102384, step = 50800 (3.172 sec)
I0804 19:53:57.586403 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6639
I0804 19:53:57.587818 140200711067520 basic_session_run_hooks.py:260] loss = 1.1377907, step = 50900 (3.158 sec)
I0804 19:54:00.680351 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 51000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:54:00.964649 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:54:01.012020 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.1916
I0804 19:54:01.013129 140200711067520 basic_session_run_hooks.py:260] loss = 1.164564, step = 51000 (3.425 sec)
I0804 19:54:04.178538 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5809
I0804 19:54:04.179876 140200711067520 basic_session_run_hooks.py:260] loss = 1.1054587, step = 51100 (3.167 sec)
I0804 19:54:07.314398 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8888
I0804 19:54:07.315788 140200711067520 basic_session_run_hooks.py:260] loss = 1.0583663, step = 51200 (3.136 sec)
I0804 19:54:10.448649 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9057
I0804 19:54:10.449871 140200711067520 basic_session_run_hooks.py:260] loss = 1.1223209, step = 51300 (3.134 sec)
I0804 19:54:13.607793 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6543
I0804 19:54:13.608874 140200711067520 basic_session_run_hooks.py:260] loss = 1.1808815, step = 51400 (3.159 sec)
I0804 19:54:16.750149 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8232
I0804 19:54:16.751516 140200711067520 basic_session_run_hooks.py:260] loss = 1.1135195, step = 51500 (3.143 sec)
I0804 19:54:19.901066 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7369
I0804 19:54:19.902602 140200711067520 basic_session_run_hooks.py:260] loss = 1.1426946, step = 51600 (3.151 sec)
I0804 19:54:23.039519 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8629
I0804 19:54:23.040961 140200711067520 basic_session_run_hooks.py:260] loss = 1.0928975, step = 51700 (3.138 sec)
I0804 19:54:26.182787 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8137
I0804 19:54:26.183879 140200711067520 basic_session_run_hooks.py:260] loss = 1.1732086, step = 51800 (3.143 sec)
I0804 19:54:29.298323 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0979
I0804 19:54:29.299856 140200711067520 basic_session_run_hooks.py:260] loss = 1.1170661, step = 51900 (3.116 sec)
I0804 19:54:32.390118 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 52000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:54:32.679187 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:54:32.716495 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.2549
I0804 19:54:32.717678 140200711067520 basic_session_run_hooks.py:260] loss = 1.1693146, step = 52000 (3.418 sec)
I0804 19:54:35.883304 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5774
I0804 19:54:35.884670 140200711067520 basic_session_run_hooks.py:260] loss = 1.0919673, step = 52100 (3.167 sec)
I0804 19:54:39.068015 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4001
I0804 19:54:39.069315 140200711067520 basic_session_run_hooks.py:260] loss = 1.1486325, step = 52200 (3.185 sec)
I0804 19:54:42.231596 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.61
I0804 19:54:42.233313 140200711067520 basic_session_run_hooks.py:260] loss = 1.0326444, step = 52300 (3.164 sec)
I0804 19:54:45.431683 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2488
I0804 19:54:45.433272 140200711067520 basic_session_run_hooks.py:260] loss = 1.1561646, step = 52400 (3.200 sec)
I0804 19:54:48.617242 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3918
I0804 19:54:48.618931 140200711067520 basic_session_run_hooks.py:260] loss = 1.1593627, step = 52500 (3.186 sec)
I0804 19:54:51.784730 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5707
I0804 19:54:51.786341 140200711067520 basic_session_run_hooks.py:260] loss = 1.096503, step = 52600 (3.167 sec)
I0804 19:54:54.933558 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.758
I0804 19:54:54.934953 140200711067520 basic_session_run_hooks.py:260] loss = 1.1462494, step = 52700 (3.149 sec)
I0804 19:54:58.086586 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7157
I0804 19:54:58.087816 140200711067520 basic_session_run_hooks.py:260] loss = 1.1285942, step = 52800 (3.153 sec)
I0804 19:55:01.256705 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5442
I0804 19:55:01.258167 140200711067520 basic_session_run_hooks.py:260] loss = 1.0621536, step = 52900 (3.170 sec)
I0804 19:55:04.366623 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 53000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:55:04.646145 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:55:04.691712 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.1118
I0804 19:55:04.692725 140200711067520 basic_session_run_hooks.py:260] loss = 1.0658259, step = 53000 (3.435 sec)
I0804 19:55:07.854251 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6204
I0804 19:55:07.855698 140200711067520 basic_session_run_hooks.py:260] loss = 1.239122, step = 53100 (3.163 sec)
I0804 19:55:11.002407 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7647
I0804 19:55:11.004044 140200711067520 basic_session_run_hooks.py:260] loss = 1.0317576, step = 53200 (3.148 sec)
I0804 19:55:14.155816 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7116
I0804 19:55:14.157128 140200711067520 basic_session_run_hooks.py:260] loss = 1.1620356, step = 53300 (3.153 sec)
I0804 19:55:17.312117 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6827
I0804 19:55:17.313552 140200711067520 basic_session_run_hooks.py:260] loss = 1.0933932, step = 53400 (3.156 sec)
I0804 19:55:20.485606 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5113
I0804 19:55:20.486971 140200711067520 basic_session_run_hooks.py:260] loss = 1.0403192, step = 53500 (3.173 sec)
I0804 19:55:23.661749 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4843
I0804 19:55:23.662831 140200711067520 basic_session_run_hooks.py:260] loss = 1.1505634, step = 53600 (3.176 sec)
I0804 19:55:26.803408 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8304
I0804 19:55:26.804740 140200711067520 basic_session_run_hooks.py:260] loss = 1.1294134, step = 53700 (3.142 sec)
I0804 19:55:29.931383 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9695
I0804 19:55:29.933047 140200711067520 basic_session_run_hooks.py:260] loss = 1.0780337, step = 53800 (3.128 sec)
I0804 19:55:33.079072 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7696
I0804 19:55:33.080737 140200711067520 basic_session_run_hooks.py:260] loss = 1.1401592, step = 53900 (3.148 sec)
I0804 19:55:36.188865 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 54000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:55:36.473089 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:55:36.518766 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0721
I0804 19:55:36.519917 140200711067520 basic_session_run_hooks.py:260] loss = 1.1109818, step = 54000 (3.439 sec)
I0804 19:55:39.651938 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9166
I0804 19:55:39.653703 140200711067520 basic_session_run_hooks.py:260] loss = 1.1557012, step = 54100 (3.134 sec)
I0804 19:55:42.756994 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2058
I0804 19:55:42.758057 140200711067520 basic_session_run_hooks.py:260] loss = 1.0555997, step = 54200 (3.104 sec)
I0804 19:55:45.889070 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9277
I0804 19:55:45.890573 140200711067520 basic_session_run_hooks.py:260] loss = 1.0703788, step = 54300 (3.133 sec)
I0804 19:55:49.073341 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4043
I0804 19:55:49.074833 140200711067520 basic_session_run_hooks.py:260] loss = 1.0740724, step = 54400 (3.184 sec)
I0804 19:55:52.229365 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6855
I0804 19:55:52.230489 140200711067520 basic_session_run_hooks.py:260] loss = 1.1376364, step = 54500 (3.156 sec)
I0804 19:55:55.378294 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7568
I0804 19:55:55.379486 140200711067520 basic_session_run_hooks.py:260] loss = 1.1044346, step = 54600 (3.149 sec)
I0804 19:55:58.572577 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3061
I0804 19:55:58.574032 140200711067520 basic_session_run_hooks.py:260] loss = 0.9625174, step = 54700 (3.195 sec)
I0804 19:56:01.772163 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2539
I0804 19:56:01.773534 140200711067520 basic_session_run_hooks.py:260] loss = 1.1996288, step = 54800 (3.200 sec)
I0804 19:56:04.949277 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.475
I0804 19:56:04.950679 140200711067520 basic_session_run_hooks.py:260] loss = 1.056813, step = 54900 (3.177 sec)
I0804 19:56:08.101654 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 55000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:56:08.418706 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 19:56:08.420932 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 19:56:08.572386 140200711067520 estimator.py:1145] Calling model_fn.
I0804 19:56:08.573518 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 19:56:08.573923 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 19:56:08.574017 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 19:56:08.574097 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 19:56:08.574162 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 19:56:08.574251 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 19:56:08.574318 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 19:56:08.667471 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 19:56:08.724606 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 19:56:08.865501 140200711067520 t2t_model.py:2172] Building model body
I0804 19:56:09.832279 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 19:56:10.553351 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 19:56:10.571363 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T19:56:10Z
I0804 19:56:10.744755 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 19:56:10.745492: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:56:10.746131: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 19:56:10.746253: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 19:56:10.746281: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 19:56:10.746319: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 19:56:10.746344: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 19:56:10.746368: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 19:56:10.746389: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 19:56:10.746412: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 19:56:10.746550: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:56:10.746996: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:56:10.747381: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 19:56:10.747438: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 19:56:10.747453: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 19:56:10.747464: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 19:56:10.747762: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:56:10.748169: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 19:56:10.748543: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 19:56:10.750221 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-55000
I0804 19:56:10.968080 140200711067520 session_manager.py:500] Running local_init_op.
I0804 19:56:11.013839 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 19:56:17.050319 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 19:56:22.395787 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 19:56:27.744117 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 19:56:33.078585 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 19:56:38.402147 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 19:56:43.762751 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 19:56:49.114289 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 19:56:54.486138 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 19:56:59.810226 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 19:57:04.649225 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-19:57:04
I0804 19:57:04.649498 140200711067520 estimator.py:2039] Saving dict for global step 55000: global_step = 55000, loss = 1.2081723, metrics-paper_generation_problem/targets/accuracy = 0.6652576, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8795241, metrics-paper_generation_problem/targets/approx_bleu_score = 0.47883806, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.2082133, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5751814, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.68875223
I0804 19:57:04.650125 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 55000: experiment/transformer/transformer_small/output/model.ckpt-55000
I0804 19:57:04.706001 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.67345
I0804 19:57:04.707399 140200711067520 basic_session_run_hooks.py:260] loss = 1.0607003, step = 55000 (59.757 sec)
I0804 19:57:07.923402 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0814
I0804 19:57:07.924884 140200711067520 basic_session_run_hooks.py:260] loss = 1.1108562, step = 55100 (3.217 sec)
I0804 19:57:11.121412 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2693
I0804 19:57:11.122961 140200711067520 basic_session_run_hooks.py:260] loss = 1.1038703, step = 55200 (3.198 sec)
I0804 19:57:14.252345 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9395
I0804 19:57:14.253849 140200711067520 basic_session_run_hooks.py:260] loss = 1.1483694, step = 55300 (3.131 sec)
I0804 19:57:17.408575 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6837
I0804 19:57:17.409916 140200711067520 basic_session_run_hooks.py:260] loss = 1.071307, step = 55400 (3.156 sec)
I0804 19:57:20.598969 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3437
I0804 19:57:20.600068 140200711067520 basic_session_run_hooks.py:260] loss = 0.99535394, step = 55500 (3.190 sec)
I0804 19:57:23.782686 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4099
I0804 19:57:23.784085 140200711067520 basic_session_run_hooks.py:260] loss = 1.0324261, step = 55600 (3.184 sec)
I0804 19:57:26.991226 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1668
I0804 19:57:26.992703 140200711067520 basic_session_run_hooks.py:260] loss = 1.076366, step = 55700 (3.209 sec)
I0804 19:57:30.197532 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1887
I0804 19:57:30.199038 140200711067520 basic_session_run_hooks.py:260] loss = 1.186993, step = 55800 (3.206 sec)
I0804 19:57:33.412509 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1044
I0804 19:57:33.414023 140200711067520 basic_session_run_hooks.py:260] loss = 0.9782503, step = 55900 (3.215 sec)
I0804 19:57:36.597505 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 56000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:57:36.886758 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:57:36.926680 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4559
I0804 19:57:36.927844 140200711067520 basic_session_run_hooks.py:260] loss = 1.1130954, step = 56000 (3.514 sec)
I0804 19:57:40.086695 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6457
I0804 19:57:40.088151 140200711067520 basic_session_run_hooks.py:260] loss = 1.0823654, step = 56100 (3.160 sec)
I0804 19:57:43.269677 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4171
I0804 19:57:43.271018 140200711067520 basic_session_run_hooks.py:260] loss = 1.1019537, step = 56200 (3.183 sec)
I0804 19:57:46.426124 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6814
I0804 19:57:46.427522 140200711067520 basic_session_run_hooks.py:260] loss = 1.055721, step = 56300 (3.157 sec)
I0804 19:57:49.552592 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9849
I0804 19:57:49.554058 140200711067520 basic_session_run_hooks.py:260] loss = 1.1056371, step = 56400 (3.127 sec)
I0804 19:57:52.722284 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5487
I0804 19:57:52.723756 140200711067520 basic_session_run_hooks.py:260] loss = 1.1686088, step = 56500 (3.170 sec)
I0804 19:57:55.863209 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8378
I0804 19:57:55.864525 140200711067520 basic_session_run_hooks.py:260] loss = 1.0882874, step = 56600 (3.141 sec)
I0804 19:57:59.021404 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6635
I0804 19:57:59.022526 140200711067520 basic_session_run_hooks.py:260] loss = 1.1754758, step = 56700 (3.158 sec)
I0804 19:58:02.176584 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6942
I0804 19:58:02.178018 140200711067520 basic_session_run_hooks.py:260] loss = 1.0952232, step = 56800 (3.155 sec)
I0804 19:58:05.317117 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8415
I0804 19:58:05.318641 140200711067520 basic_session_run_hooks.py:260] loss = 1.0726768, step = 56900 (3.141 sec)
I0804 19:58:08.451125 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 57000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:58:08.740071 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:58:08.782262 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8588
I0804 19:58:08.783478 140200711067520 basic_session_run_hooks.py:260] loss = 1.0740817, step = 57000 (3.465 sec)
I0804 19:58:11.923342 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8363
I0804 19:58:11.924623 140200711067520 basic_session_run_hooks.py:260] loss = 1.1353008, step = 57100 (3.141 sec)
I0804 19:58:15.066680 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8134
I0804 19:58:15.067974 140200711067520 basic_session_run_hooks.py:260] loss = 1.087218, step = 57200 (3.143 sec)
I0804 19:58:18.234451 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5681
I0804 19:58:18.235657 140200711067520 basic_session_run_hooks.py:260] loss = 1.0135503, step = 57300 (3.168 sec)
I0804 19:58:21.400079 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5892
I0804 19:58:21.401508 140200711067520 basic_session_run_hooks.py:260] loss = 1.078632, step = 57400 (3.166 sec)
I0804 19:58:24.546097 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7862
I0804 19:58:24.547499 140200711067520 basic_session_run_hooks.py:260] loss = 1.1145657, step = 57500 (3.146 sec)
I0804 19:58:27.672161 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9891
I0804 19:58:27.673565 140200711067520 basic_session_run_hooks.py:260] loss = 1.1039739, step = 57600 (3.126 sec)
I0804 19:58:30.846071 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5069
I0804 19:58:30.847771 140200711067520 basic_session_run_hooks.py:260] loss = 1.1326375, step = 57700 (3.174 sec)
I0804 19:58:34.006285 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6434
I0804 19:58:34.007808 140200711067520 basic_session_run_hooks.py:260] loss = 1.0811329, step = 57800 (3.160 sec)
I0804 19:58:37.174202 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5664
I0804 19:58:37.175614 140200711067520 basic_session_run_hooks.py:260] loss = 1.0591686, step = 57900 (3.168 sec)
I0804 19:58:40.308157 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 58000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:58:40.603532 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:58:40.645732 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8059
I0804 19:58:40.646864 140200711067520 basic_session_run_hooks.py:260] loss = 1.0997509, step = 58000 (3.471 sec)
I0804 19:58:43.834755 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3574
I0804 19:58:43.836171 140200711067520 basic_session_run_hooks.py:260] loss = 1.1451458, step = 58100 (3.189 sec)
I0804 19:58:47.004729 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5465
I0804 19:58:47.006628 140200711067520 basic_session_run_hooks.py:260] loss = 1.0974491, step = 58200 (3.170 sec)
I0804 19:58:50.179342 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4993
I0804 19:58:50.180735 140200711067520 basic_session_run_hooks.py:260] loss = 1.0874507, step = 58300 (3.174 sec)
I0804 19:58:53.350754 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5317
I0804 19:58:53.352098 140200711067520 basic_session_run_hooks.py:260] loss = 1.073808, step = 58400 (3.171 sec)
I0804 19:58:56.525976 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4944
I0804 19:58:56.527580 140200711067520 basic_session_run_hooks.py:260] loss = 1.1622556, step = 58500 (3.175 sec)
I0804 19:58:59.712857 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3785
I0804 19:58:59.714261 140200711067520 basic_session_run_hooks.py:260] loss = 1.0765821, step = 58600 (3.187 sec)
I0804 19:59:02.843840 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9386
I0804 19:59:02.845301 140200711067520 basic_session_run_hooks.py:260] loss = 1.1775134, step = 58700 (3.131 sec)
I0804 19:59:06.096846 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7408
I0804 19:59:06.099079 140200711067520 basic_session_run_hooks.py:260] loss = 1.0854988, step = 58800 (3.254 sec)
I0804 19:59:09.290861 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3086
I0804 19:59:09.292299 140200711067520 basic_session_run_hooks.py:260] loss = 1.1573375, step = 58900 (3.193 sec)
I0804 19:59:12.469443 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 59000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:59:12.756153 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:59:12.794199 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5439
I0804 19:59:12.795320 140200711067520 basic_session_run_hooks.py:260] loss = 0.9799618, step = 59000 (3.503 sec)
I0804 19:59:15.984203 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3485
I0804 19:59:15.985927 140200711067520 basic_session_run_hooks.py:260] loss = 1.0810544, step = 59100 (3.191 sec)
I0804 19:59:19.171921 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.37
I0804 19:59:19.172963 140200711067520 basic_session_run_hooks.py:260] loss = 1.1105347, step = 59200 (3.187 sec)
I0804 19:59:22.376674 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2037
I0804 19:59:22.377999 140200711067520 basic_session_run_hooks.py:260] loss = 1.1356585, step = 59300 (3.205 sec)
I0804 19:59:25.569039 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3248
I0804 19:59:25.570461 140200711067520 basic_session_run_hooks.py:260] loss = 1.1243539, step = 59400 (3.192 sec)
I0804 19:59:28.742250 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5137
I0804 19:59:28.743693 140200711067520 basic_session_run_hooks.py:260] loss = 1.1171454, step = 59500 (3.173 sec)
I0804 19:59:31.929295 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.377
I0804 19:59:31.930773 140200711067520 basic_session_run_hooks.py:260] loss = 1.1176981, step = 59600 (3.187 sec)
I0804 19:59:35.115385 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3863
I0804 19:59:35.116705 140200711067520 basic_session_run_hooks.py:260] loss = 1.084817, step = 59700 (3.186 sec)
I0804 19:59:38.307304 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3293
I0804 19:59:38.308831 140200711067520 basic_session_run_hooks.py:260] loss = 1.0568261, step = 59800 (3.192 sec)
I0804 19:59:41.500842 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3131
I0804 19:59:41.502188 140200711067520 basic_session_run_hooks.py:260] loss = 1.129903, step = 59900 (3.193 sec)
I0804 19:59:44.630290 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 60000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 19:59:44.914000 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 19:59:44.957900 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.9262
I0804 19:59:44.958976 140200711067520 basic_session_run_hooks.py:260] loss = 1.1291734, step = 60000 (3.457 sec)
I0804 19:59:48.151537 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3126
I0804 19:59:48.152798 140200711067520 basic_session_run_hooks.py:260] loss = 1.0978385, step = 60100 (3.194 sec)
I0804 19:59:51.334220 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.42
I0804 19:59:51.335705 140200711067520 basic_session_run_hooks.py:260] loss = 1.0864881, step = 60200 (3.183 sec)
I0804 19:59:54.518456 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4047
I0804 19:59:54.519983 140200711067520 basic_session_run_hooks.py:260] loss = 1.1259824, step = 60300 (3.184 sec)
I0804 19:59:57.741552 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0262
I0804 19:59:57.742829 140200711067520 basic_session_run_hooks.py:260] loss = 0.98768294, step = 60400 (3.223 sec)
I0804 20:00:00.914269 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5184
I0804 20:00:00.915325 140200711067520 basic_session_run_hooks.py:260] loss = 1.1623096, step = 60500 (3.172 sec)
I0804 20:00:04.110997 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.282
I0804 20:00:04.112062 140200711067520 basic_session_run_hooks.py:260] loss = 1.1329099, step = 60600 (3.197 sec)
I0804 20:00:07.313582 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2249
I0804 20:00:07.314976 140200711067520 basic_session_run_hooks.py:260] loss = 1.0935318, step = 60700 (3.203 sec)
I0804 20:00:10.514258 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2434
I0804 20:00:10.515748 140200711067520 basic_session_run_hooks.py:260] loss = 1.0428413, step = 60800 (3.201 sec)
I0804 20:00:13.723223 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1627
I0804 20:00:13.724742 140200711067520 basic_session_run_hooks.py:260] loss = 1.0652039, step = 60900 (3.209 sec)
I0804 20:00:16.889145 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 61000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:00:17.190969 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:00:17.237255 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4571
I0804 20:00:17.238502 140200711067520 basic_session_run_hooks.py:260] loss = 1.0390912, step = 61000 (3.514 sec)
I0804 20:00:20.439202 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2312
I0804 20:00:20.440581 140200711067520 basic_session_run_hooks.py:260] loss = 1.0524248, step = 61100 (3.202 sec)
I0804 20:00:23.606970 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5681
I0804 20:00:23.608496 140200711067520 basic_session_run_hooks.py:260] loss = 1.0799774, step = 61200 (3.168 sec)
I0804 20:00:26.782703 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4887
I0804 20:00:26.784242 140200711067520 basic_session_run_hooks.py:260] loss = 1.1131146, step = 61300 (3.175 sec)
I0804 20:00:29.936334 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7094
I0804 20:00:29.937651 140200711067520 basic_session_run_hooks.py:260] loss = 1.0296532, step = 61400 (3.154 sec)
I0804 20:00:33.064540 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9675
I0804 20:00:33.065941 140200711067520 basic_session_run_hooks.py:260] loss = 1.0813007, step = 61500 (3.128 sec)
I0804 20:00:36.217468 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7166
I0804 20:00:36.218871 140200711067520 basic_session_run_hooks.py:260] loss = 1.0611889, step = 61600 (3.153 sec)
I0804 20:00:39.364645 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7743
I0804 20:00:39.366012 140200711067520 basic_session_run_hooks.py:260] loss = 1.1264815, step = 61700 (3.147 sec)
I0804 20:00:42.497057 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9242
I0804 20:00:42.498326 140200711067520 basic_session_run_hooks.py:260] loss = 1.016971, step = 61800 (3.132 sec)
I0804 20:00:45.674835 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4685
I0804 20:00:45.676022 140200711067520 basic_session_run_hooks.py:260] loss = 1.1872503, step = 61900 (3.178 sec)
I0804 20:00:48.857390 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 62000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:00:49.162325 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:00:49.197671 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3861
I0804 20:00:49.198752 140200711067520 basic_session_run_hooks.py:260] loss = 1.1432111, step = 62000 (3.523 sec)
I0804 20:00:52.381360 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4104
I0804 20:00:52.382855 140200711067520 basic_session_run_hooks.py:260] loss = 1.1526952, step = 62100 (3.184 sec)
I0804 20:00:55.547927 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5799
I0804 20:00:55.549347 140200711067520 basic_session_run_hooks.py:260] loss = 1.0113653, step = 62200 (3.166 sec)
I0804 20:00:58.747536 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2539
I0804 20:00:58.748919 140200711067520 basic_session_run_hooks.py:260] loss = 1.2043588, step = 62300 (3.200 sec)
I0804 20:01:01.937005 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3531
I0804 20:01:01.938518 140200711067520 basic_session_run_hooks.py:260] loss = 1.116093, step = 62400 (3.190 sec)
I0804 20:01:05.132124 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2976
I0804 20:01:05.133593 140200711067520 basic_session_run_hooks.py:260] loss = 1.108949, step = 62500 (3.195 sec)
I0804 20:01:08.329405 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2768
I0804 20:01:08.330923 140200711067520 basic_session_run_hooks.py:260] loss = 1.0415306, step = 62600 (3.197 sec)
I0804 20:01:11.537076 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1754
I0804 20:01:11.538225 140200711067520 basic_session_run_hooks.py:260] loss = 1.0223739, step = 62700 (3.207 sec)
I0804 20:01:14.747886 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1448
I0804 20:01:14.749020 140200711067520 basic_session_run_hooks.py:260] loss = 1.0807977, step = 62800 (3.211 sec)
I0804 20:01:17.946460 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2637
I0804 20:01:17.947859 140200711067520 basic_session_run_hooks.py:260] loss = 1.1336738, step = 62900 (3.199 sec)
I0804 20:01:21.135985 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 63000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:01:21.439476 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:01:21.477395 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3208
I0804 20:01:21.478518 140200711067520 basic_session_run_hooks.py:260] loss = 1.1521982, step = 63000 (3.531 sec)
I0804 20:01:24.668403 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3383
I0804 20:01:24.669788 140200711067520 basic_session_run_hooks.py:260] loss = 1.0666636, step = 63100 (3.191 sec)
I0804 20:01:27.839080 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5391
I0804 20:01:27.840449 140200711067520 basic_session_run_hooks.py:260] loss = 1.0431764, step = 63200 (3.171 sec)
I0804 20:01:31.040694 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2343
I0804 20:01:31.041808 140200711067520 basic_session_run_hooks.py:260] loss = 1.0743295, step = 63300 (3.201 sec)
I0804 20:01:34.218746 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4656
I0804 20:01:34.220015 140200711067520 basic_session_run_hooks.py:260] loss = 1.0895246, step = 63400 (3.178 sec)
I0804 20:01:37.437317 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0699
I0804 20:01:37.439094 140200711067520 basic_session_run_hooks.py:260] loss = 0.9844716, step = 63500 (3.219 sec)
I0804 20:01:40.610623 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5129
I0804 20:01:40.612129 140200711067520 basic_session_run_hooks.py:260] loss = 1.1071937, step = 63600 (3.173 sec)
I0804 20:01:43.775626 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5954
I0804 20:01:43.777091 140200711067520 basic_session_run_hooks.py:260] loss = 1.1737705, step = 63700 (3.165 sec)
I0804 20:01:46.952267 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4797
I0804 20:01:46.953696 140200711067520 basic_session_run_hooks.py:260] loss = 0.9817141, step = 63800 (3.177 sec)
I0804 20:01:50.115278 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6155
I0804 20:01:50.116620 140200711067520 basic_session_run_hooks.py:260] loss = 1.1641625, step = 63900 (3.163 sec)
I0804 20:01:53.253710 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 64000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:01:53.535241 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:01:53.575854 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8969
I0804 20:01:53.576944 140200711067520 basic_session_run_hooks.py:260] loss = 1.0819494, step = 64000 (3.460 sec)
I0804 20:01:56.743225 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5722
I0804 20:01:56.744640 140200711067520 basic_session_run_hooks.py:260] loss = 1.0828749, step = 64100 (3.168 sec)
I0804 20:01:59.899181 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.686
I0804 20:01:59.900770 140200711067520 basic_session_run_hooks.py:260] loss = 1.1087382, step = 64200 (3.156 sec)
I0804 20:02:03.088401 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3558
I0804 20:02:03.089808 140200711067520 basic_session_run_hooks.py:260] loss = 1.1634599, step = 64300 (3.189 sec)
I0804 20:02:06.276986 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.362
I0804 20:02:06.278599 140200711067520 basic_session_run_hooks.py:260] loss = 1.1525248, step = 64400 (3.189 sec)
I0804 20:02:09.460023 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4164
I0804 20:02:09.461511 140200711067520 basic_session_run_hooks.py:260] loss = 1.1218172, step = 64500 (3.183 sec)
I0804 20:02:12.636127 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.485
I0804 20:02:12.637626 140200711067520 basic_session_run_hooks.py:260] loss = 1.0896468, step = 64600 (3.176 sec)
I0804 20:02:15.824991 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3594
I0804 20:02:15.826500 140200711067520 basic_session_run_hooks.py:260] loss = 1.0977443, step = 64700 (3.189 sec)
I0804 20:02:19.011084 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3863
I0804 20:02:19.012395 140200711067520 basic_session_run_hooks.py:260] loss = 1.0793471, step = 64800 (3.186 sec)
I0804 20:02:22.189590 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4615
I0804 20:02:22.190901 140200711067520 basic_session_run_hooks.py:260] loss = 1.2049052, step = 64900 (3.179 sec)
I0804 20:02:25.344986 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 65000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:02:25.649158 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:02:25.687772 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5859
I0804 20:02:25.688845 140200711067520 basic_session_run_hooks.py:260] loss = 1.1213692, step = 65000 (3.498 sec)
I0804 20:02:28.904155 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.091
I0804 20:02:28.905311 140200711067520 basic_session_run_hooks.py:260] loss = 1.0770715, step = 65100 (3.216 sec)
I0804 20:02:32.115595 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1388
I0804 20:02:32.116798 140200711067520 basic_session_run_hooks.py:260] loss = 1.1015029, step = 65200 (3.211 sec)
I0804 20:02:35.297253 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4301
I0804 20:02:35.298624 140200711067520 basic_session_run_hooks.py:260] loss = 1.1646515, step = 65300 (3.182 sec)
I0804 20:02:38.491969 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3015
I0804 20:02:38.493090 140200711067520 basic_session_run_hooks.py:260] loss = 1.1056373, step = 65400 (3.194 sec)
I0804 20:02:41.693200 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2382
I0804 20:02:41.694595 140200711067520 basic_session_run_hooks.py:260] loss = 1.1140419, step = 65500 (3.202 sec)
I0804 20:02:44.904713 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1382
I0804 20:02:44.905979 140200711067520 basic_session_run_hooks.py:260] loss = 1.0323123, step = 65600 (3.211 sec)
I0804 20:02:48.131568 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9899
I0804 20:02:48.132721 140200711067520 basic_session_run_hooks.py:260] loss = 1.1968182, step = 65700 (3.227 sec)
I0804 20:02:51.334278 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2233
I0804 20:02:51.335843 140200711067520 basic_session_run_hooks.py:260] loss = 1.0720929, step = 65800 (3.203 sec)
I0804 20:02:54.556981 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.03
I0804 20:02:54.558207 140200711067520 basic_session_run_hooks.py:260] loss = 1.0762043, step = 65900 (3.222 sec)
I0804 20:02:57.730866 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 66000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:02:58.010131 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:02:58.058901 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5557
I0804 20:02:58.060156 140200711067520 basic_session_run_hooks.py:260] loss = 1.2130185, step = 66000 (3.502 sec)
I0804 20:03:01.255877 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2795
I0804 20:03:01.257112 140200711067520 basic_session_run_hooks.py:260] loss = 1.089858, step = 66100 (3.197 sec)
I0804 20:03:04.446989 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3372
I0804 20:03:04.448528 140200711067520 basic_session_run_hooks.py:260] loss = 1.0309135, step = 66200 (3.191 sec)
I0804 20:03:07.653712 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1843
I0804 20:03:07.655115 140200711067520 basic_session_run_hooks.py:260] loss = 1.129274, step = 66300 (3.207 sec)
I0804 20:03:10.870346 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0885
I0804 20:03:10.871954 140200711067520 basic_session_run_hooks.py:260] loss = 1.0420932, step = 66400 (3.217 sec)
I0804 20:03:14.073011 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.224
I0804 20:03:14.074660 140200711067520 basic_session_run_hooks.py:260] loss = 1.127213, step = 66500 (3.203 sec)
I0804 20:03:17.270124 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2782
I0804 20:03:17.271274 140200711067520 basic_session_run_hooks.py:260] loss = 1.0641245, step = 66600 (3.197 sec)
I0804 20:03:20.495810 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0011
I0804 20:03:20.497183 140200711067520 basic_session_run_hooks.py:260] loss = 1.122011, step = 66700 (3.226 sec)
I0804 20:03:23.772474 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5192
I0804 20:03:23.773743 140200711067520 basic_session_run_hooks.py:260] loss = 1.1796178, step = 66800 (3.277 sec)
I0804 20:03:26.993255 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.048
I0804 20:03:26.994363 140200711067520 basic_session_run_hooks.py:260] loss = 1.048032, step = 66900 (3.221 sec)
I0804 20:03:30.200612 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 67000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:03:30.484068 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:03:30.525152 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3135
I0804 20:03:30.526464 140200711067520 basic_session_run_hooks.py:260] loss = 1.1662976, step = 67000 (3.532 sec)
I0804 20:03:33.743602 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.071
I0804 20:03:33.744930 140200711067520 basic_session_run_hooks.py:260] loss = 1.1361343, step = 67100 (3.218 sec)
I0804 20:03:36.954260 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1461
I0804 20:03:36.955943 140200711067520 basic_session_run_hooks.py:260] loss = 1.1209587, step = 67200 (3.211 sec)
I0804 20:03:40.164288 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1527
I0804 20:03:40.165779 140200711067520 basic_session_run_hooks.py:260] loss = 1.139861, step = 67300 (3.210 sec)
I0804 20:03:43.402733 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8786
I0804 20:03:43.404397 140200711067520 basic_session_run_hooks.py:260] loss = 1.0599569, step = 67400 (3.239 sec)
I0804 20:03:46.595015 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3257
I0804 20:03:46.596325 140200711067520 basic_session_run_hooks.py:260] loss = 1.0427226, step = 67500 (3.192 sec)
I0804 20:03:49.805303 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1498
I0804 20:03:49.806813 140200711067520 basic_session_run_hooks.py:260] loss = 1.1284019, step = 67600 (3.210 sec)
I0804 20:03:53.028898 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0214
I0804 20:03:53.030411 140200711067520 basic_session_run_hooks.py:260] loss = 1.1699252, step = 67700 (3.224 sec)
I0804 20:03:56.214610 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3901
I0804 20:03:56.216149 140200711067520 basic_session_run_hooks.py:260] loss = 1.1088475, step = 67800 (3.186 sec)
I0804 20:03:59.434250 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0595
I0804 20:03:59.435794 140200711067520 basic_session_run_hooks.py:260] loss = 1.0738709, step = 67900 (3.220 sec)
I0804 20:04:02.629070 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 68000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:04:02.938549 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:04:02.977609 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2216
I0804 20:04:02.978956 140200711067520 basic_session_run_hooks.py:260] loss = 1.0249337, step = 68000 (3.543 sec)
I0804 20:04:06.251870 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5414
I0804 20:04:06.253051 140200711067520 basic_session_run_hooks.py:260] loss = 1.1434551, step = 68100 (3.274 sec)
I0804 20:04:09.524607 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.556
I0804 20:04:09.525952 140200711067520 basic_session_run_hooks.py:260] loss = 1.1150979, step = 68200 (3.273 sec)
I0804 20:04:12.768445 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8278
I0804 20:04:12.769786 140200711067520 basic_session_run_hooks.py:260] loss = 1.0820508, step = 68300 (3.244 sec)
I0804 20:04:16.011986 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.83
I0804 20:04:16.013285 140200711067520 basic_session_run_hooks.py:260] loss = 1.1342075, step = 68400 (3.243 sec)
I0804 20:04:19.223104 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1417
I0804 20:04:19.224442 140200711067520 basic_session_run_hooks.py:260] loss = 1.1852944, step = 68500 (3.211 sec)
I0804 20:04:22.453696 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.954
I0804 20:04:22.455214 140200711067520 basic_session_run_hooks.py:260] loss = 1.1395863, step = 68600 (3.231 sec)
I0804 20:04:25.736100 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4656
I0804 20:04:25.737309 140200711067520 basic_session_run_hooks.py:260] loss = 1.1588401, step = 68700 (3.282 sec)
I0804 20:04:28.941397 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1984
I0804 20:04:28.942833 140200711067520 basic_session_run_hooks.py:260] loss = 1.1182827, step = 68800 (3.206 sec)
I0804 20:04:32.121184 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4485
I0804 20:04:32.122637 140200711067520 basic_session_run_hooks.py:260] loss = 1.0335522, step = 68900 (3.180 sec)
I0804 20:04:35.264096 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 69000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:04:35.557720 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:04:35.597970 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7621
I0804 20:04:35.599231 140200711067520 basic_session_run_hooks.py:260] loss = 1.1805687, step = 69000 (3.477 sec)
I0804 20:04:38.764920 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5762
I0804 20:04:38.766111 140200711067520 basic_session_run_hooks.py:260] loss = 1.1233019, step = 69100 (3.167 sec)
I0804 20:04:41.915025 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7452
I0804 20:04:41.916404 140200711067520 basic_session_run_hooks.py:260] loss = 1.1023625, step = 69200 (3.150 sec)
I0804 20:04:45.074327 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6524
I0804 20:04:45.075852 140200711067520 basic_session_run_hooks.py:260] loss = 1.0813107, step = 69300 (3.159 sec)
I0804 20:04:48.251136 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4783
I0804 20:04:48.252785 140200711067520 basic_session_run_hooks.py:260] loss = 1.1801779, step = 69400 (3.177 sec)
I0804 20:04:51.483151 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9404
I0804 20:04:51.484842 140200711067520 basic_session_run_hooks.py:260] loss = 1.1497302, step = 69500 (3.232 sec)
I0804 20:04:54.683652 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2451
I0804 20:04:54.685040 140200711067520 basic_session_run_hooks.py:260] loss = 1.1299773, step = 69600 (3.200 sec)
I0804 20:04:57.916701 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9306
I0804 20:04:57.918310 140200711067520 basic_session_run_hooks.py:260] loss = 1.1902096, step = 69700 (3.233 sec)
I0804 20:05:01.161183 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8215
I0804 20:05:01.162302 140200711067520 basic_session_run_hooks.py:260] loss = 1.1872562, step = 69800 (3.244 sec)
I0804 20:05:04.388344 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9873
I0804 20:05:04.389863 140200711067520 basic_session_run_hooks.py:260] loss = 1.1199487, step = 69900 (3.228 sec)
I0804 20:05:07.591885 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 70000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:05:08.169645 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:05:08.209684 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 26.1685
I0804 20:05:08.210872 140200711067520 basic_session_run_hooks.py:260] loss = 1.0819657, step = 70000 (3.821 sec)
I0804 20:05:11.421261 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1375
I0804 20:05:11.422674 140200711067520 basic_session_run_hooks.py:260] loss = 1.1229033, step = 70100 (3.212 sec)
I0804 20:05:14.625521 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2087
I0804 20:05:14.626640 140200711067520 basic_session_run_hooks.py:260] loss = 1.1517715, step = 70200 (3.204 sec)
I0804 20:05:17.863830 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8838
I0804 20:05:17.865860 140200711067520 basic_session_run_hooks.py:260] loss = 1.169714, step = 70300 (3.239 sec)
I0804 20:05:21.100519 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8922
I0804 20:05:21.101827 140200711067520 basic_session_run_hooks.py:260] loss = 1.0721549, step = 70400 (3.236 sec)
I0804 20:05:24.355759 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7195
I0804 20:05:24.357021 140200711067520 basic_session_run_hooks.py:260] loss = 1.0594697, step = 70500 (3.255 sec)
I0804 20:05:27.603896 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.787
I0804 20:05:27.605541 140200711067520 basic_session_run_hooks.py:260] loss = 1.1690818, step = 70600 (3.249 sec)
I0804 20:05:30.877451 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5478
I0804 20:05:30.878770 140200711067520 basic_session_run_hooks.py:260] loss = 1.1125445, step = 70700 (3.273 sec)
I0804 20:05:34.135004 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6979
I0804 20:05:34.136182 140200711067520 basic_session_run_hooks.py:260] loss = 1.1462021, step = 70800 (3.257 sec)
I0804 20:05:37.362625 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9825
I0804 20:05:37.364140 140200711067520 basic_session_run_hooks.py:260] loss = 1.0332291, step = 70900 (3.228 sec)
I0804 20:05:40.571511 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 71000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:05:40.862532 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:05:40.907778 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2075
I0804 20:05:40.908920 140200711067520 basic_session_run_hooks.py:260] loss = 1.0362604, step = 71000 (3.545 sec)
I0804 20:05:44.134382 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9925
I0804 20:05:44.135887 140200711067520 basic_session_run_hooks.py:260] loss = 1.0330987, step = 71100 (3.227 sec)
I0804 20:05:47.307942 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5104
I0804 20:05:47.309498 140200711067520 basic_session_run_hooks.py:260] loss = 1.087853, step = 71200 (3.174 sec)
I0804 20:05:50.471940 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6055
I0804 20:05:50.473402 140200711067520 basic_session_run_hooks.py:260] loss = 1.0309798, step = 71300 (3.164 sec)
I0804 20:05:53.643441 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5309
I0804 20:05:53.645060 140200711067520 basic_session_run_hooks.py:260] loss = 1.0561467, step = 71400 (3.172 sec)
I0804 20:05:56.829612 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3858
I0804 20:05:56.831239 140200711067520 basic_session_run_hooks.py:260] loss = 1.052373, step = 71500 (3.186 sec)
I0804 20:06:00.001786 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5238
I0804 20:06:00.003138 140200711067520 basic_session_run_hooks.py:260] loss = 1.0607759, step = 71600 (3.172 sec)
I0804 20:06:03.174648 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5174
I0804 20:06:03.175943 140200711067520 basic_session_run_hooks.py:260] loss = 1.1104323, step = 71700 (3.173 sec)
I0804 20:06:06.334080 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6512
I0804 20:06:06.335737 140200711067520 basic_session_run_hooks.py:260] loss = 1.1178298, step = 71800 (3.160 sec)
I0804 20:06:09.542638 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1666
I0804 20:06:09.544119 140200711067520 basic_session_run_hooks.py:260] loss = 1.0931473, step = 71900 (3.208 sec)
I0804 20:06:12.706974 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 72000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:06:13.015545 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 20:06:13.017395 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 20:06:13.166236 140200711067520 estimator.py:1145] Calling model_fn.
I0804 20:06:13.167335 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 20:06:13.167768 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 20:06:13.167863 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 20:06:13.167945 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 20:06:13.168012 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 20:06:13.168093 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 20:06:13.168160 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 20:06:13.259334 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 20:06:13.315537 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 20:06:13.455307 140200711067520 t2t_model.py:2172] Building model body
I0804 20:06:14.154498 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 20:06:15.069193 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 20:06:15.087389 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T20:06:15Z
I0804 20:06:15.254745 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 20:06:15.255369: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:06:15.255786: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 20:06:15.255897: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 20:06:15.255922: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 20:06:15.255947: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 20:06:15.255971: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 20:06:15.255992: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 20:06:15.256014: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 20:06:15.256038: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 20:06:15.256139: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:06:15.256544: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:06:15.256862: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 20:06:15.256905: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 20:06:15.256919: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 20:06:15.256929: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 20:06:15.257205: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:06:15.257603: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:06:15.257934: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 20:06:15.259507 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-72000
I0804 20:06:15.462068 140200711067520 session_manager.py:500] Running local_init_op.
I0804 20:06:15.505507 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 20:06:21.544710 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 20:06:26.924376 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 20:06:32.292227 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 20:06:37.650861 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 20:06:43.016377 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 20:06:48.458147 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 20:06:53.879169 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 20:06:59.282540 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 20:07:04.637601 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 20:07:09.521304 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-20:07:09
I0804 20:07:09.521582 140200711067520 estimator.py:2039] Saving dict for global step 72000: global_step = 72000, loss = 1.1989051, metrics-paper_generation_problem/targets/accuracy = 0.6678424, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8812666, metrics-paper_generation_problem/targets/approx_bleu_score = 0.48139358, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1989408, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5775634, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.69063985
I0804 20:07:09.522156 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 72000: experiment/transformer/transformer_small/output/model.ckpt-72000
I0804 20:07:09.576997 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.66571
I0804 20:07:09.577971 140200711067520 basic_session_run_hooks.py:260] loss = 1.0907108, step = 72000 (60.034 sec)
I0804 20:07:12.814450 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8891
I0804 20:07:12.816033 140200711067520 basic_session_run_hooks.py:260] loss = 1.0396234, step = 72100 (3.238 sec)
I0804 20:07:15.978399 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6058
I0804 20:07:15.979861 140200711067520 basic_session_run_hooks.py:260] loss = 1.0632346, step = 72200 (3.164 sec)
I0804 20:07:19.137922 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6503
I0804 20:07:19.139643 140200711067520 basic_session_run_hooks.py:260] loss = 1.1170584, step = 72300 (3.160 sec)
I0804 20:07:22.318974 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4362
I0804 20:07:22.320518 140200711067520 basic_session_run_hooks.py:260] loss = 1.1483496, step = 72400 (3.181 sec)
I0804 20:07:25.511006 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3279
I0804 20:07:25.513089 140200711067520 basic_session_run_hooks.py:260] loss = 1.1461112, step = 72500 (3.193 sec)
I0804 20:07:28.693078 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.426
I0804 20:07:28.694827 140200711067520 basic_session_run_hooks.py:260] loss = 1.0507752, step = 72600 (3.182 sec)
I0804 20:07:31.948198 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7209
I0804 20:07:31.949704 140200711067520 basic_session_run_hooks.py:260] loss = 1.0992683, step = 72700 (3.255 sec)
I0804 20:07:35.218333 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5797
I0804 20:07:35.219944 140200711067520 basic_session_run_hooks.py:260] loss = 1.0968764, step = 72800 (3.270 sec)
I0804 20:07:38.430293 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1338
I0804 20:07:38.431918 140200711067520 basic_session_run_hooks.py:260] loss = 1.0191457, step = 72900 (3.212 sec)
I0804 20:07:41.599009 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 73000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:07:41.880611 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:07:41.929053 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5814
I0804 20:07:41.930195 140200711067520 basic_session_run_hooks.py:260] loss = 1.0360324, step = 73000 (3.498 sec)
I0804 20:07:45.145836 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.087
I0804 20:07:45.146989 140200711067520 basic_session_run_hooks.py:260] loss = 1.0842419, step = 73100 (3.217 sec)
I0804 20:07:48.343397 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2738
I0804 20:07:48.344935 140200711067520 basic_session_run_hooks.py:260] loss = 1.08971, step = 73200 (3.198 sec)
I0804 20:07:51.555562 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.132
I0804 20:07:51.557040 140200711067520 basic_session_run_hooks.py:260] loss = 1.0601825, step = 73300 (3.212 sec)
I0804 20:07:54.790749 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9097
I0804 20:07:54.792184 140200711067520 basic_session_run_hooks.py:260] loss = 1.0535579, step = 73400 (3.235 sec)
I0804 20:07:58.011956 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0444
I0804 20:07:58.013109 140200711067520 basic_session_run_hooks.py:260] loss = 1.0951271, step = 73500 (3.221 sec)
I0804 20:08:01.259564 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7934
I0804 20:08:01.261498 140200711067520 basic_session_run_hooks.py:260] loss = 1.1121389, step = 73600 (3.248 sec)
I0804 20:08:04.462906 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2157
I0804 20:08:04.464245 140200711067520 basic_session_run_hooks.py:260] loss = 1.0704823, step = 73700 (3.203 sec)
I0804 20:08:07.679458 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0894
I0804 20:08:07.680971 140200711067520 basic_session_run_hooks.py:260] loss = 1.2339379, step = 73800 (3.217 sec)
I0804 20:08:10.902520 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0264
I0804 20:08:10.903895 140200711067520 basic_session_run_hooks.py:260] loss = 1.068816, step = 73900 (3.223 sec)
I0804 20:08:14.037501 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 74000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:08:14.316738 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:08:14.372550 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.818
I0804 20:08:14.373661 140200711067520 basic_session_run_hooks.py:260] loss = 1.0687195, step = 74000 (3.470 sec)
I0804 20:08:17.541458 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5568
I0804 20:08:17.542830 140200711067520 basic_session_run_hooks.py:260] loss = 1.0764691, step = 74100 (3.169 sec)
I0804 20:08:20.729772 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3644
I0804 20:08:20.731458 140200711067520 basic_session_run_hooks.py:260] loss = 1.1537442, step = 74200 (3.189 sec)
I0804 20:08:23.880363 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.74
I0804 20:08:23.881705 140200711067520 basic_session_run_hooks.py:260] loss = 1.1549251, step = 74300 (3.150 sec)
I0804 20:08:27.045293 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5963
I0804 20:08:27.046740 140200711067520 basic_session_run_hooks.py:260] loss = 1.1959239, step = 74400 (3.165 sec)
I0804 20:08:30.208306 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6156
I0804 20:08:30.209788 140200711067520 basic_session_run_hooks.py:260] loss = 1.102683, step = 74500 (3.163 sec)
I0804 20:08:33.360254 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7263
I0804 20:08:33.361594 140200711067520 basic_session_run_hooks.py:260] loss = 1.0943588, step = 74600 (3.152 sec)
I0804 20:08:36.506150 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7875
I0804 20:08:36.507722 140200711067520 basic_session_run_hooks.py:260] loss = 1.0980246, step = 74700 (3.146 sec)
I0804 20:08:39.634251 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9682
I0804 20:08:39.635643 140200711067520 basic_session_run_hooks.py:260] loss = 1.0765951, step = 74800 (3.128 sec)
I0804 20:08:42.822220 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3681
I0804 20:08:42.823719 140200711067520 basic_session_run_hooks.py:260] loss = 1.1634231, step = 74900 (3.188 sec)
I0804 20:08:45.984337 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 75000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:08:46.263197 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:08:46.303474 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7251
I0804 20:08:46.304603 140200711067520 basic_session_run_hooks.py:260] loss = 1.1216342, step = 75000 (3.481 sec)
I0804 20:08:49.520185 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0878
I0804 20:08:49.521412 140200711067520 basic_session_run_hooks.py:260] loss = 1.105029, step = 75100 (3.217 sec)
I0804 20:08:52.728658 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1673
I0804 20:08:52.729822 140200711067520 basic_session_run_hooks.py:260] loss = 1.0110593, step = 75200 (3.208 sec)
I0804 20:08:55.932696 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2107
I0804 20:08:55.934054 140200711067520 basic_session_run_hooks.py:260] loss = 1.0768695, step = 75300 (3.204 sec)
I0804 20:08:59.142473 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1548
I0804 20:08:59.143941 140200711067520 basic_session_run_hooks.py:260] loss = 1.0921804, step = 75400 (3.210 sec)
I0804 20:09:02.340386 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2703
I0804 20:09:02.341754 140200711067520 basic_session_run_hooks.py:260] loss = 1.1395296, step = 75500 (3.198 sec)
I0804 20:09:05.573868 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9264
I0804 20:09:05.575073 140200711067520 basic_session_run_hooks.py:260] loss = 1.1084011, step = 75600 (3.233 sec)
I0804 20:09:08.803269 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9656
I0804 20:09:08.804530 140200711067520 basic_session_run_hooks.py:260] loss = 1.0538913, step = 75700 (3.229 sec)
I0804 20:09:12.048762 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.812
I0804 20:09:12.050293 140200711067520 basic_session_run_hooks.py:260] loss = 1.0612044, step = 75800 (3.246 sec)
I0804 20:09:15.267514 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0682
I0804 20:09:15.268817 140200711067520 basic_session_run_hooks.py:260] loss = 1.070931, step = 75900 (3.219 sec)
I0804 20:09:18.454203 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 76000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:09:18.745639 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:09:18.786675 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4156
I0804 20:09:18.787736 140200711067520 basic_session_run_hooks.py:260] loss = 1.052507, step = 76000 (3.519 sec)
I0804 20:09:22.018296 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9442
I0804 20:09:22.020030 140200711067520 basic_session_run_hooks.py:260] loss = 1.268545, step = 76100 (3.232 sec)
I0804 20:09:25.246020 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9817
I0804 20:09:25.247088 140200711067520 basic_session_run_hooks.py:260] loss = 1.0596001, step = 76200 (3.227 sec)
I0804 20:09:28.400205 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.704
I0804 20:09:28.401734 140200711067520 basic_session_run_hooks.py:260] loss = 1.1017228, step = 76300 (3.155 sec)
I0804 20:09:31.586620 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.383
I0804 20:09:31.588308 140200711067520 basic_session_run_hooks.py:260] loss = 1.0865198, step = 76400 (3.187 sec)
I0804 20:09:34.774572 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3684
I0804 20:09:34.775980 140200711067520 basic_session_run_hooks.py:260] loss = 1.1628367, step = 76500 (3.188 sec)
I0804 20:09:37.944199 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5492
I0804 20:09:37.945525 140200711067520 basic_session_run_hooks.py:260] loss = 1.040666, step = 76600 (3.170 sec)
I0804 20:09:41.080647 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8832
I0804 20:09:41.081718 140200711067520 basic_session_run_hooks.py:260] loss = 1.044273, step = 76700 (3.136 sec)
I0804 20:09:44.233734 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7151
I0804 20:09:44.234975 140200711067520 basic_session_run_hooks.py:260] loss = 1.1335367, step = 76800 (3.153 sec)
I0804 20:09:47.393936 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6435
I0804 20:09:47.395269 140200711067520 basic_session_run_hooks.py:260] loss = 1.1006494, step = 76900 (3.160 sec)
I0804 20:09:50.515280 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 77000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:09:50.798906 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:09:50.832820 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0791
I0804 20:09:50.833884 140200711067520 basic_session_run_hooks.py:260] loss = 1.0611411, step = 77000 (3.439 sec)
I0804 20:09:53.993814 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6356
I0804 20:09:53.995081 140200711067520 basic_session_run_hooks.py:260] loss = 1.1222262, step = 77100 (3.161 sec)
I0804 20:09:57.122617 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9612
I0804 20:09:57.124214 140200711067520 basic_session_run_hooks.py:260] loss = 1.125504, step = 77200 (3.129 sec)
I0804 20:10:00.288473 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5871
I0804 20:10:00.289828 140200711067520 basic_session_run_hooks.py:260] loss = 1.1718837, step = 77300 (3.166 sec)
I0804 20:10:03.442084 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7094
I0804 20:10:03.443448 140200711067520 basic_session_run_hooks.py:260] loss = 0.99264693, step = 77400 (3.154 sec)
I0804 20:10:06.620172 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4656
I0804 20:10:06.621360 140200711067520 basic_session_run_hooks.py:260] loss = 1.0638542, step = 77500 (3.178 sec)
I0804 20:10:09.821240 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2397
I0804 20:10:09.822684 140200711067520 basic_session_run_hooks.py:260] loss = 1.0959909, step = 77600 (3.201 sec)
I0804 20:10:12.991404 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5439
I0804 20:10:12.992727 140200711067520 basic_session_run_hooks.py:260] loss = 1.1320714, step = 77700 (3.170 sec)
I0804 20:10:16.178178 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3797
I0804 20:10:16.179754 140200711067520 basic_session_run_hooks.py:260] loss = 1.1780677, step = 77800 (3.187 sec)
I0804 20:10:19.361253 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4162
I0804 20:10:19.362690 140200711067520 basic_session_run_hooks.py:260] loss = 1.105932, step = 77900 (3.183 sec)
I0804 20:10:22.530811 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 78000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:10:22.829528 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:10:22.868410 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5128
I0804 20:10:22.869798 140200711067520 basic_session_run_hooks.py:260] loss = 1.0340463, step = 78000 (3.507 sec)
I0804 20:10:26.049177 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4393
I0804 20:10:26.050611 140200711067520 basic_session_run_hooks.py:260] loss = 1.1419394, step = 78100 (3.181 sec)
I0804 20:10:29.208091 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6565
I0804 20:10:29.209515 140200711067520 basic_session_run_hooks.py:260] loss = 1.1477963, step = 78200 (3.159 sec)
I0804 20:10:32.394014 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3879
I0804 20:10:32.395148 140200711067520 basic_session_run_hooks.py:260] loss = 1.1646553, step = 78300 (3.186 sec)
I0804 20:10:35.573658 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4501
I0804 20:10:35.574784 140200711067520 basic_session_run_hooks.py:260] loss = 1.159935, step = 78400 (3.180 sec)
I0804 20:10:38.731767 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6645
I0804 20:10:38.732951 140200711067520 basic_session_run_hooks.py:260] loss = 1.0654843, step = 78500 (3.158 sec)
I0804 20:10:41.975352 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.83
I0804 20:10:41.976678 140200711067520 basic_session_run_hooks.py:260] loss = 1.0266618, step = 78600 (3.244 sec)
I0804 20:10:45.139549 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6039
I0804 20:10:45.141689 140200711067520 basic_session_run_hooks.py:260] loss = 1.0819777, step = 78700 (3.165 sec)
I0804 20:10:48.295812 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6829
I0804 20:10:48.297722 140200711067520 basic_session_run_hooks.py:260] loss = 1.039888, step = 78800 (3.156 sec)
I0804 20:10:51.467771 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5261
I0804 20:10:51.468990 140200711067520 basic_session_run_hooks.py:260] loss = 1.1037085, step = 78900 (3.171 sec)
I0804 20:10:54.556910 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 79000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:10:54.844910 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:10:54.883035 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.2801
I0804 20:10:54.884028 140200711067520 basic_session_run_hooks.py:260] loss = 1.0690082, step = 79000 (3.415 sec)
I0804 20:10:58.033400 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7425
I0804 20:10:58.034677 140200711067520 basic_session_run_hooks.py:260] loss = 1.0806563, step = 79100 (3.151 sec)
I0804 20:11:01.182159 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7587
I0804 20:11:01.183516 140200711067520 basic_session_run_hooks.py:260] loss = 1.0443734, step = 79200 (3.149 sec)
I0804 20:11:04.328861 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7792
I0804 20:11:04.330054 140200711067520 basic_session_run_hooks.py:260] loss = 0.99003565, step = 79300 (3.147 sec)
I0804 20:11:07.539768 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1439
I0804 20:11:07.541083 140200711067520 basic_session_run_hooks.py:260] loss = 1.0867827, step = 79400 (3.211 sec)
I0804 20:11:10.747927 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1705
I0804 20:11:10.749314 140200711067520 basic_session_run_hooks.py:260] loss = 1.0948111, step = 79500 (3.208 sec)
I0804 20:11:13.930756 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4187
I0804 20:11:13.932219 140200711067520 basic_session_run_hooks.py:260] loss = 1.1152786, step = 79600 (3.183 sec)
I0804 20:11:17.105663 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4969
I0804 20:11:17.107043 140200711067520 basic_session_run_hooks.py:260] loss = 1.1001859, step = 79700 (3.175 sec)
I0804 20:11:20.297049 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3344
I0804 20:11:20.298565 140200711067520 basic_session_run_hooks.py:260] loss = 1.0200825, step = 79800 (3.192 sec)
I0804 20:11:23.514610 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0794
I0804 20:11:23.515942 140200711067520 basic_session_run_hooks.py:260] loss = 1.1355283, step = 79900 (3.217 sec)
I0804 20:11:26.705556 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 80000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:11:26.987870 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:11:27.033118 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4209
I0804 20:11:27.034369 140200711067520 basic_session_run_hooks.py:260] loss = 1.0151666, step = 80000 (3.518 sec)
I0804 20:11:30.241466 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1691
I0804 20:11:30.242828 140200711067520 basic_session_run_hooks.py:260] loss = 1.0851172, step = 80100 (3.208 sec)
I0804 20:11:33.462212 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0487
I0804 20:11:33.463683 140200711067520 basic_session_run_hooks.py:260] loss = 1.0469815, step = 80200 (3.221 sec)
I0804 20:11:36.631934 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5484
I0804 20:11:36.633215 140200711067520 basic_session_run_hooks.py:260] loss = 1.0398356, step = 80300 (3.170 sec)
I0804 20:11:39.783549 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.73
I0804 20:11:39.785150 140200711067520 basic_session_run_hooks.py:260] loss = 1.1708634, step = 80400 (3.152 sec)
I0804 20:11:42.945322 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6276
I0804 20:11:42.946746 140200711067520 basic_session_run_hooks.py:260] loss = 1.0279952, step = 80500 (3.162 sec)
I0804 20:11:46.121274 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4866
I0804 20:11:46.122701 140200711067520 basic_session_run_hooks.py:260] loss = 1.0672946, step = 80600 (3.176 sec)
I0804 20:11:49.289620 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5622
I0804 20:11:49.291023 140200711067520 basic_session_run_hooks.py:260] loss = 1.1954333, step = 80700 (3.168 sec)
I0804 20:11:52.505453 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0962
I0804 20:11:52.506795 140200711067520 basic_session_run_hooks.py:260] loss = 1.0884513, step = 80800 (3.216 sec)
I0804 20:11:55.686620 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4349
I0804 20:11:55.687974 140200711067520 basic_session_run_hooks.py:260] loss = 1.0563899, step = 80900 (3.181 sec)
I0804 20:11:58.893840 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 81000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:11:59.173819 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:11:59.220635 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2964
I0804 20:11:59.221691 140200711067520 basic_session_run_hooks.py:260] loss = 1.1311381, step = 81000 (3.534 sec)
I0804 20:12:02.369358 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.759
I0804 20:12:02.370915 140200711067520 basic_session_run_hooks.py:260] loss = 1.134732, step = 81100 (3.149 sec)
I0804 20:12:05.516466 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7754
I0804 20:12:05.517878 140200711067520 basic_session_run_hooks.py:260] loss = 1.1015836, step = 81200 (3.147 sec)
I0804 20:12:08.665502 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7559
I0804 20:12:08.666906 140200711067520 basic_session_run_hooks.py:260] loss = 1.1614529, step = 81300 (3.149 sec)
I0804 20:12:11.820689 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6936
I0804 20:12:11.822018 140200711067520 basic_session_run_hooks.py:260] loss = 1.0516775, step = 81400 (3.155 sec)
I0804 20:12:14.979306 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6596
I0804 20:12:14.980780 140200711067520 basic_session_run_hooks.py:260] loss = 1.2540115, step = 81500 (3.159 sec)
I0804 20:12:18.144203 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5965
I0804 20:12:18.145634 140200711067520 basic_session_run_hooks.py:260] loss = 1.0617516, step = 81600 (3.165 sec)
I0804 20:12:21.301113 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6765
I0804 20:12:21.302316 140200711067520 basic_session_run_hooks.py:260] loss = 1.1223067, step = 81700 (3.157 sec)
I0804 20:12:24.520343 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0632
I0804 20:12:24.521855 140200711067520 basic_session_run_hooks.py:260] loss = 1.1358083, step = 81800 (3.220 sec)
I0804 20:12:27.708473 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3666
I0804 20:12:27.709771 140200711067520 basic_session_run_hooks.py:260] loss = 1.0623918, step = 81900 (3.188 sec)
I0804 20:12:30.841625 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 82000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:12:31.117980 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:12:31.167535 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.9094
I0804 20:12:31.168674 140200711067520 basic_session_run_hooks.py:260] loss = 1.1335658, step = 82000 (3.459 sec)
I0804 20:12:34.359511 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3288
I0804 20:12:34.360738 140200711067520 basic_session_run_hooks.py:260] loss = 1.1010051, step = 82100 (3.192 sec)
I0804 20:12:37.541270 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.429
I0804 20:12:37.542614 140200711067520 basic_session_run_hooks.py:260] loss = 1.063753, step = 82200 (3.182 sec)
I0804 20:12:40.729081 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3696
I0804 20:12:40.730503 140200711067520 basic_session_run_hooks.py:260] loss = 1.1495227, step = 82300 (3.188 sec)
I0804 20:12:43.925177 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.288
I0804 20:12:43.926579 140200711067520 basic_session_run_hooks.py:260] loss = 1.1219604, step = 82400 (3.196 sec)
I0804 20:12:47.123904 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2625
I0804 20:12:47.125307 140200711067520 basic_session_run_hooks.py:260] loss = 1.106279, step = 82500 (3.199 sec)
I0804 20:12:50.349740 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9999
I0804 20:12:50.351184 140200711067520 basic_session_run_hooks.py:260] loss = 1.119957, step = 82600 (3.226 sec)
I0804 20:12:53.583317 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9254
I0804 20:12:53.584474 140200711067520 basic_session_run_hooks.py:260] loss = 1.0921786, step = 82700 (3.233 sec)
I0804 20:12:56.812773 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.965
I0804 20:12:56.814286 140200711067520 basic_session_run_hooks.py:260] loss = 1.1015517, step = 82800 (3.230 sec)
I0804 20:12:59.991860 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4555
I0804 20:12:59.993241 140200711067520 basic_session_run_hooks.py:260] loss = 1.1107757, step = 82900 (3.179 sec)
I0804 20:13:03.147395 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 83000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:13:03.439115 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:13:03.483621 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6388
I0804 20:13:03.484768 140200711067520 basic_session_run_hooks.py:260] loss = 1.0449158, step = 83000 (3.492 sec)
I0804 20:13:06.712357 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9718
I0804 20:13:06.713670 140200711067520 basic_session_run_hooks.py:260] loss = 1.123624, step = 83100 (3.229 sec)
I0804 20:13:09.925607 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1212
I0804 20:13:09.926815 140200711067520 basic_session_run_hooks.py:260] loss = 1.0995291, step = 83200 (3.213 sec)
I0804 20:13:13.157571 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9411
I0804 20:13:13.158698 140200711067520 basic_session_run_hooks.py:260] loss = 1.094116, step = 83300 (3.232 sec)
I0804 20:13:16.314133 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.68
I0804 20:13:16.315347 140200711067520 basic_session_run_hooks.py:260] loss = 0.9826397, step = 83400 (3.157 sec)
I0804 20:13:19.493439 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4536
I0804 20:13:19.494658 140200711067520 basic_session_run_hooks.py:260] loss = 0.9793355, step = 83500 (3.179 sec)
I0804 20:13:22.650063 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6793
I0804 20:13:22.651454 140200711067520 basic_session_run_hooks.py:260] loss = 1.0177397, step = 83600 (3.157 sec)
I0804 20:13:25.814441 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.602
I0804 20:13:25.815687 140200711067520 basic_session_run_hooks.py:260] loss = 1.0649003, step = 83700 (3.164 sec)
I0804 20:13:28.988583 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5044
I0804 20:13:28.989957 140200711067520 basic_session_run_hooks.py:260] loss = 1.1228178, step = 83800 (3.174 sec)
I0804 20:13:32.165070 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4812
I0804 20:13:32.166249 140200711067520 basic_session_run_hooks.py:260] loss = 1.062548, step = 83900 (3.176 sec)
I0804 20:13:35.311009 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 84000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:13:35.590847 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:13:35.641044 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7688
I0804 20:13:35.642134 140200711067520 basic_session_run_hooks.py:260] loss = 1.0609404, step = 84000 (3.476 sec)
I0804 20:13:38.814723 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5091
I0804 20:13:38.815942 140200711067520 basic_session_run_hooks.py:260] loss = 0.99955326, step = 84100 (3.174 sec)
I0804 20:13:41.965512 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7384
I0804 20:13:41.966845 140200711067520 basic_session_run_hooks.py:260] loss = 1.0992888, step = 84200 (3.151 sec)
I0804 20:13:45.143091 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4702
I0804 20:13:45.144782 140200711067520 basic_session_run_hooks.py:260] loss = 1.1736614, step = 84300 (3.178 sec)
I0804 20:13:48.312406 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5528
I0804 20:13:48.313758 140200711067520 basic_session_run_hooks.py:260] loss = 1.1385373, step = 84400 (3.169 sec)
I0804 20:13:51.506627 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3064
I0804 20:13:51.508165 140200711067520 basic_session_run_hooks.py:260] loss = 1.0920149, step = 84500 (3.194 sec)
I0804 20:13:54.676172 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5503
I0804 20:13:54.677338 140200711067520 basic_session_run_hooks.py:260] loss = 1.0977134, step = 84600 (3.169 sec)
I0804 20:13:57.852408 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4838
I0804 20:13:57.853874 140200711067520 basic_session_run_hooks.py:260] loss = 1.0700604, step = 84700 (3.177 sec)
I0804 20:14:01.053218 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.242
I0804 20:14:01.054621 140200711067520 basic_session_run_hooks.py:260] loss = 1.0089452, step = 84800 (3.201 sec)
I0804 20:14:04.283352 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9585
I0804 20:14:04.284754 140200711067520 basic_session_run_hooks.py:260] loss = 1.0580257, step = 84900 (3.230 sec)
I0804 20:14:07.444360 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 85000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:14:07.719208 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:14:07.764919 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7226
I0804 20:14:07.765900 140200711067520 basic_session_run_hooks.py:260] loss = 1.0506088, step = 85000 (3.481 sec)
I0804 20:14:10.983354 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0711
I0804 20:14:10.984768 140200711067520 basic_session_run_hooks.py:260] loss = 1.1109627, step = 85100 (3.219 sec)
I0804 20:14:14.155729 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5222
I0804 20:14:14.157010 140200711067520 basic_session_run_hooks.py:260] loss = 1.120176, step = 85200 (3.172 sec)
I0804 20:14:17.337169 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4323
I0804 20:14:17.338741 140200711067520 basic_session_run_hooks.py:260] loss = 1.1474797, step = 85300 (3.182 sec)
I0804 20:14:20.541961 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2032
I0804 20:14:20.543462 140200711067520 basic_session_run_hooks.py:260] loss = 1.0356333, step = 85400 (3.205 sec)
I0804 20:14:23.733611 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3319
I0804 20:14:23.734874 140200711067520 basic_session_run_hooks.py:260] loss = 1.1477603, step = 85500 (3.191 sec)
I0804 20:14:26.984311 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7626
I0804 20:14:26.985723 140200711067520 basic_session_run_hooks.py:260] loss = 0.9538162, step = 85600 (3.251 sec)
I0804 20:14:30.216172 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9419
I0804 20:14:30.217473 140200711067520 basic_session_run_hooks.py:260] loss = 1.1257819, step = 85700 (3.232 sec)
I0804 20:14:33.390700 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5006
I0804 20:14:33.392061 140200711067520 basic_session_run_hooks.py:260] loss = 1.0887583, step = 85800 (3.175 sec)
I0804 20:14:36.538681 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7663
I0804 20:14:36.540055 140200711067520 basic_session_run_hooks.py:260] loss = 1.0835884, step = 85900 (3.148 sec)
I0804 20:14:39.640685 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 86000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:14:39.925746 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:14:39.962512 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.2072
I0804 20:14:39.963721 140200711067520 basic_session_run_hooks.py:260] loss = 1.0997707, step = 86000 (3.424 sec)
I0804 20:14:43.087143 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0039
I0804 20:14:43.088365 140200711067520 basic_session_run_hooks.py:260] loss = 1.0970075, step = 86100 (3.125 sec)
I0804 20:14:46.233095 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7867
I0804 20:14:46.234395 140200711067520 basic_session_run_hooks.py:260] loss = 1.1053895, step = 86200 (3.146 sec)
I0804 20:14:49.364665 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9329
I0804 20:14:49.366007 140200711067520 basic_session_run_hooks.py:260] loss = 1.0896249, step = 86300 (3.132 sec)
I0804 20:14:52.517271 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7198
I0804 20:14:52.518796 140200711067520 basic_session_run_hooks.py:260] loss = 1.0708487, step = 86400 (3.153 sec)
I0804 20:14:55.679184 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6263
I0804 20:14:55.680438 140200711067520 basic_session_run_hooks.py:260] loss = 1.127845, step = 86500 (3.162 sec)
I0804 20:14:58.838172 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6559
I0804 20:14:58.839734 140200711067520 basic_session_run_hooks.py:260] loss = 1.0742633, step = 86600 (3.159 sec)
I0804 20:15:02.018173 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4463
I0804 20:15:02.019474 140200711067520 basic_session_run_hooks.py:260] loss = 1.0624479, step = 86700 (3.180 sec)
I0804 20:15:05.186121 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5663
I0804 20:15:05.187664 140200711067520 basic_session_run_hooks.py:260] loss = 1.1226318, step = 86800 (3.168 sec)
I0804 20:15:08.367748 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4305
I0804 20:15:08.369099 140200711067520 basic_session_run_hooks.py:260] loss = 1.0737425, step = 86900 (3.181 sec)
I0804 20:15:11.492783 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 87000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:15:11.768696 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:15:11.814259 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0148
I0804 20:15:11.815684 140200711067520 basic_session_run_hooks.py:260] loss = 1.0445493, step = 87000 (3.447 sec)
I0804 20:15:14.966458 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.724
I0804 20:15:14.967796 140200711067520 basic_session_run_hooks.py:260] loss = 1.0745836, step = 87100 (3.152 sec)
I0804 20:15:18.141044 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5
I0804 20:15:18.142478 140200711067520 basic_session_run_hooks.py:260] loss = 1.0215164, step = 87200 (3.175 sec)
I0804 20:15:21.319917 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4579
I0804 20:15:21.321438 140200711067520 basic_session_run_hooks.py:260] loss = 1.0350939, step = 87300 (3.179 sec)
I0804 20:15:24.514174 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.306
I0804 20:15:24.515631 140200711067520 basic_session_run_hooks.py:260] loss = 1.0461724, step = 87400 (3.194 sec)
I0804 20:15:27.707524 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3152
I0804 20:15:27.708675 140200711067520 basic_session_run_hooks.py:260] loss = 1.0850618, step = 87500 (3.193 sec)
I0804 20:15:30.900570 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3181
I0804 20:15:30.901972 140200711067520 basic_session_run_hooks.py:260] loss = 1.1929997, step = 87600 (3.193 sec)
I0804 20:15:34.079940 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4526
I0804 20:15:34.081181 140200711067520 basic_session_run_hooks.py:260] loss = 1.0528797, step = 87700 (3.179 sec)
I0804 20:15:37.274648 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3017
I0804 20:15:37.276038 140200711067520 basic_session_run_hooks.py:260] loss = 1.1205204, step = 87800 (3.195 sec)
I0804 20:15:40.457277 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4207
I0804 20:15:40.458765 140200711067520 basic_session_run_hooks.py:260] loss = 1.1561316, step = 87900 (3.183 sec)
I0804 20:15:43.637196 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 88000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:15:43.911596 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:15:43.958920 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5579
I0804 20:15:43.959962 140200711067520 basic_session_run_hooks.py:260] loss = 1.0728935, step = 88000 (3.501 sec)
I0804 20:15:47.115763 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6773
I0804 20:15:47.117117 140200711067520 basic_session_run_hooks.py:260] loss = 1.0964166, step = 88100 (3.157 sec)
I0804 20:15:50.279376 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6095
I0804 20:15:50.280884 140200711067520 basic_session_run_hooks.py:260] loss = 1.135759, step = 88200 (3.164 sec)
I0804 20:15:53.435809 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6816
I0804 20:15:53.437071 140200711067520 basic_session_run_hooks.py:260] loss = 1.1431156, step = 88300 (3.156 sec)
I0804 20:15:56.531709 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.3005
I0804 20:15:56.533304 140200711067520 basic_session_run_hooks.py:260] loss = 1.0846504, step = 88400 (3.096 sec)
I0804 20:15:59.771984 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8617
I0804 20:15:59.773725 140200711067520 basic_session_run_hooks.py:260] loss = 1.0312486, step = 88500 (3.240 sec)
I0804 20:16:02.918168 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7845
I0804 20:16:02.919581 140200711067520 basic_session_run_hooks.py:260] loss = 1.1277521, step = 88600 (3.146 sec)
I0804 20:16:06.049751 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9329
I0804 20:16:06.051068 140200711067520 basic_session_run_hooks.py:260] loss = 1.0761828, step = 88700 (3.131 sec)
I0804 20:16:09.211054 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6323
I0804 20:16:09.212410 140200711067520 basic_session_run_hooks.py:260] loss = 1.0875453, step = 88800 (3.161 sec)
I0804 20:16:12.346281 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8956
I0804 20:16:12.348454 140200711067520 basic_session_run_hooks.py:260] loss = 1.0572449, step = 88900 (3.136 sec)
I0804 20:16:15.486899 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 89000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:16:15.790226 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 20:16:15.791605 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 20:16:15.942974 140200711067520 estimator.py:1145] Calling model_fn.
I0804 20:16:15.943973 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 20:16:15.944383 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 20:16:15.944491 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 20:16:15.944575 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 20:16:15.944642 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 20:16:15.944725 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 20:16:15.944791 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 20:16:16.034520 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 20:16:16.097321 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 20:16:16.496737 140200711067520 t2t_model.py:2172] Building model body
I0804 20:16:17.188823 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 20:16:18.115863 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 20:16:18.134031 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T20:16:18Z
I0804 20:16:18.298622 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 20:16:18.299257: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:16:18.299696: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 20:16:18.299796: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 20:16:18.299822: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 20:16:18.299842: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 20:16:18.299861: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 20:16:18.299885: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 20:16:18.299912: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 20:16:18.299938: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 20:16:18.300042: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:16:18.300456: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:16:18.300772: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 20:16:18.300814: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 20:16:18.300828: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 20:16:18.300838: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 20:16:18.301117: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:16:18.301524: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:16:18.301851: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 20:16:18.303308 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-89000
I0804 20:16:18.506633 140200711067520 session_manager.py:500] Running local_init_op.
I0804 20:16:18.551559 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 20:16:24.603571 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 20:16:29.946759 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 20:16:35.328732 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 20:16:40.694717 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 20:16:46.101931 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 20:16:51.531584 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 20:16:56.898134 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 20:17:02.346850 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 20:17:07.750754 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 20:17:12.697595 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-20:17:12
I0804 20:17:12.697836 140200711067520 estimator.py:2039] Saving dict for global step 89000: global_step = 89000, loss = 1.1894685, metrics-paper_generation_problem/targets/accuracy = 0.6700534, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8822958, metrics-paper_generation_problem/targets/approx_bleu_score = 0.483857, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1895083, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.57913834, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.6918579
I0804 20:17:12.698331 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 89000: experiment/transformer/transformer_small/output/model.ckpt-89000
I0804 20:17:12.753583 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.65543
I0804 20:17:12.754903 140200711067520 basic_session_run_hooks.py:260] loss = 1.0770298, step = 89000 (60.406 sec)
I0804 20:17:15.963360 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1552
I0804 20:17:15.965071 140200711067520 basic_session_run_hooks.py:260] loss = 1.0734535, step = 89100 (3.210 sec)
I0804 20:17:19.134585 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.534
I0804 20:17:19.135993 140200711067520 basic_session_run_hooks.py:260] loss = 1.0490459, step = 89200 (3.171 sec)
I0804 20:17:22.344174 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1563
I0804 20:17:22.345804 140200711067520 basic_session_run_hooks.py:260] loss = 1.1544073, step = 89300 (3.210 sec)
I0804 20:17:25.532686 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3623
I0804 20:17:25.533826 140200711067520 basic_session_run_hooks.py:260] loss = 1.0372403, step = 89400 (3.188 sec)
I0804 20:17:28.751262 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0697
I0804 20:17:28.752789 140200711067520 basic_session_run_hooks.py:260] loss = 1.106136, step = 89500 (3.219 sec)
I0804 20:17:31.998407 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7963
I0804 20:17:31.999795 140200711067520 basic_session_run_hooks.py:260] loss = 1.0685325, step = 89600 (3.247 sec)
I0804 20:17:35.230858 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9365
I0804 20:17:35.232148 140200711067520 basic_session_run_hooks.py:260] loss = 1.0577533, step = 89700 (3.232 sec)
I0804 20:17:38.439021 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1705
I0804 20:17:38.440816 140200711067520 basic_session_run_hooks.py:260] loss = 1.0584471, step = 89800 (3.209 sec)
I0804 20:17:41.671962 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9313
I0804 20:17:41.673055 140200711067520 basic_session_run_hooks.py:260] loss = 1.0381382, step = 89900 (3.232 sec)
I0804 20:17:44.858505 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 90000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:17:45.153625 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:17:45.201305 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3338
I0804 20:17:45.202508 140200711067520 basic_session_run_hooks.py:260] loss = 1.1205088, step = 90000 (3.529 sec)
I0804 20:17:48.464438 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6457
I0804 20:17:48.465872 140200711067520 basic_session_run_hooks.py:260] loss = 1.1207095, step = 90100 (3.263 sec)
I0804 20:17:51.675531 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1422
I0804 20:17:51.676911 140200711067520 basic_session_run_hooks.py:260] loss = 1.1304272, step = 90200 (3.211 sec)
I0804 20:17:54.886701 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.141
I0804 20:17:54.888010 140200711067520 basic_session_run_hooks.py:260] loss = 1.1350524, step = 90300 (3.211 sec)
I0804 20:17:58.134043 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7945
I0804 20:17:58.135524 140200711067520 basic_session_run_hooks.py:260] loss = 1.1025925, step = 90400 (3.248 sec)
I0804 20:18:01.339629 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1955
I0804 20:18:01.340952 140200711067520 basic_session_run_hooks.py:260] loss = 0.9940217, step = 90500 (3.205 sec)
I0804 20:18:04.541000 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2367
I0804 20:18:04.542478 140200711067520 basic_session_run_hooks.py:260] loss = 1.02123, step = 90600 (3.202 sec)
I0804 20:18:07.740451 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2555
I0804 20:18:07.741647 140200711067520 basic_session_run_hooks.py:260] loss = 1.0880059, step = 90700 (3.199 sec)
I0804 20:18:10.968882 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9746
I0804 20:18:10.970033 140200711067520 basic_session_run_hooks.py:260] loss = 1.1242177, step = 90800 (3.228 sec)
I0804 20:18:14.174378 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1967
I0804 20:18:14.175777 140200711067520 basic_session_run_hooks.py:260] loss = 1.0545102, step = 90900 (3.206 sec)
I0804 20:18:17.323189 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 91000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:18:17.625350 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:18:17.661736 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6747
I0804 20:18:17.662770 140200711067520 basic_session_run_hooks.py:260] loss = 1.0681576, step = 91000 (3.487 sec)
I0804 20:18:20.838078 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.483
I0804 20:18:20.839629 140200711067520 basic_session_run_hooks.py:260] loss = 1.0422543, step = 91100 (3.177 sec)
I0804 20:18:24.042926 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2027
I0804 20:18:24.044137 140200711067520 basic_session_run_hooks.py:260] loss = 1.1071485, step = 91200 (3.205 sec)
I0804 20:18:27.212270 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5522
I0804 20:18:27.213659 140200711067520 basic_session_run_hooks.py:260] loss = 1.0967331, step = 91300 (3.170 sec)
I0804 20:18:30.386177 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.507
I0804 20:18:30.387802 140200711067520 basic_session_run_hooks.py:260] loss = 1.1605645, step = 91400 (3.174 sec)
I0804 20:18:33.570678 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4022
I0804 20:18:33.572586 140200711067520 basic_session_run_hooks.py:260] loss = 1.0648465, step = 91500 (3.185 sec)
I0804 20:18:36.763683 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3184
I0804 20:18:36.765728 140200711067520 basic_session_run_hooks.py:260] loss = 1.1063894, step = 91600 (3.193 sec)
I0804 20:18:40.075170 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.1979
I0804 20:18:40.076537 140200711067520 basic_session_run_hooks.py:260] loss = 1.0138506, step = 91700 (3.311 sec)
I0804 20:18:43.333112 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.694
I0804 20:18:43.334814 140200711067520 basic_session_run_hooks.py:260] loss = 1.0511189, step = 91800 (3.258 sec)
I0804 20:18:46.590754 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6975
I0804 20:18:46.592364 140200711067520 basic_session_run_hooks.py:260] loss = 1.0489569, step = 91900 (3.258 sec)
I0804 20:18:49.808375 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 92000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:18:50.108719 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:18:50.154540 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.0598
I0804 20:18:50.155695 140200711067520 basic_session_run_hooks.py:260] loss = 1.0578697, step = 92000 (3.563 sec)
I0804 20:18:53.441521 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4233
I0804 20:18:53.442718 140200711067520 basic_session_run_hooks.py:260] loss = 1.1162076, step = 92100 (3.287 sec)
I0804 20:18:56.686084 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8206
I0804 20:18:56.687519 140200711067520 basic_session_run_hooks.py:260] loss = 1.0369121, step = 92200 (3.245 sec)
I0804 20:18:59.947907 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6579
I0804 20:18:59.949310 140200711067520 basic_session_run_hooks.py:260] loss = 1.1801578, step = 92300 (3.262 sec)
I0804 20:19:03.230019 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4681
I0804 20:19:03.231506 140200711067520 basic_session_run_hooks.py:260] loss = 1.0947824, step = 92400 (3.282 sec)
I0804 20:19:06.458793 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9714
I0804 20:19:06.460124 140200711067520 basic_session_run_hooks.py:260] loss = 1.0699073, step = 92500 (3.229 sec)
I0804 20:19:09.689555 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9529
I0804 20:19:09.690982 140200711067520 basic_session_run_hooks.py:260] loss = 0.98915344, step = 92600 (3.231 sec)
I0804 20:19:12.930438 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8556
I0804 20:19:12.931840 140200711067520 basic_session_run_hooks.py:260] loss = 1.0891852, step = 92700 (3.241 sec)
I0804 20:19:16.150551 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0549
I0804 20:19:16.151682 140200711067520 basic_session_run_hooks.py:260] loss = 1.0590013, step = 92800 (3.220 sec)
I0804 20:19:19.374931 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0135
I0804 20:19:19.376086 140200711067520 basic_session_run_hooks.py:260] loss = 1.0946568, step = 92900 (3.224 sec)
I0804 20:19:22.551909 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 93000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:19:22.838754 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:19:22.879962 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5303
I0804 20:19:22.880921 140200711067520 basic_session_run_hooks.py:260] loss = 1.0376781, step = 93000 (3.505 sec)
I0804 20:19:26.095354 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1007
I0804 20:19:26.096704 140200711067520 basic_session_run_hooks.py:260] loss = 1.0134411, step = 93100 (3.216 sec)
I0804 20:19:29.318773 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0231
I0804 20:19:29.320068 140200711067520 basic_session_run_hooks.py:260] loss = 1.1790397, step = 93200 (3.223 sec)
I0804 20:19:32.463184 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.802
I0804 20:19:32.464499 140200711067520 basic_session_run_hooks.py:260] loss = 1.0032008, step = 93300 (3.144 sec)
I0804 20:19:35.623244 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6453
I0804 20:19:35.624647 140200711067520 basic_session_run_hooks.py:260] loss = 1.0843843, step = 93400 (3.160 sec)
I0804 20:19:38.844577 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.043
I0804 20:19:38.845777 140200711067520 basic_session_run_hooks.py:260] loss = 1.172229, step = 93500 (3.221 sec)
I0804 20:19:42.053792 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1601
I0804 20:19:42.055332 140200711067520 basic_session_run_hooks.py:260] loss = 1.0190147, step = 93600 (3.210 sec)
I0804 20:19:45.245873 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3277
I0804 20:19:45.247098 140200711067520 basic_session_run_hooks.py:260] loss = 1.1224363, step = 93700 (3.192 sec)
I0804 20:19:48.407105 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6331
I0804 20:19:48.408475 140200711067520 basic_session_run_hooks.py:260] loss = 1.0905083, step = 93800 (3.161 sec)
I0804 20:19:51.534130 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9794
I0804 20:19:51.535684 140200711067520 basic_session_run_hooks.py:260] loss = 1.0845029, step = 93900 (3.127 sec)
I0804 20:19:54.685462 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 94000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:19:54.968299 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:19:55.009220 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.776
I0804 20:19:55.010275 140200711067520 basic_session_run_hooks.py:260] loss = 1.1340847, step = 94000 (3.475 sec)
I0804 20:19:58.176517 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5731
I0804 20:19:58.177702 140200711067520 basic_session_run_hooks.py:260] loss = 1.0994358, step = 94100 (3.167 sec)
I0804 20:20:01.352899 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.482
I0804 20:20:01.354080 140200711067520 basic_session_run_hooks.py:260] loss = 1.1026734, step = 94200 (3.176 sec)
I0804 20:20:04.527233 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5027
I0804 20:20:04.528552 140200711067520 basic_session_run_hooks.py:260] loss = 1.0766397, step = 94300 (3.174 sec)
I0804 20:20:07.695603 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5623
I0804 20:20:07.696954 140200711067520 basic_session_run_hooks.py:260] loss = 1.1367471, step = 94400 (3.168 sec)
I0804 20:20:10.909096 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1186
I0804 20:20:10.910276 140200711067520 basic_session_run_hooks.py:260] loss = 1.0830876, step = 94500 (3.213 sec)
I0804 20:20:14.160408 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7567
I0804 20:20:14.161624 140200711067520 basic_session_run_hooks.py:260] loss = 1.097334, step = 94600 (3.251 sec)
I0804 20:20:17.354995 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3031
I0804 20:20:17.356074 140200711067520 basic_session_run_hooks.py:260] loss = 1.0701654, step = 94700 (3.194 sec)
I0804 20:20:20.574191 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0636
I0804 20:20:20.575726 140200711067520 basic_session_run_hooks.py:260] loss = 1.1376612, step = 94800 (3.220 sec)
I0804 20:20:23.775578 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2366
I0804 20:20:23.777049 140200711067520 basic_session_run_hooks.py:260] loss = 1.1763451, step = 94900 (3.201 sec)
I0804 20:20:26.947929 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 95000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:20:27.235631 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:20:27.271275 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6063
I0804 20:20:27.272395 140200711067520 basic_session_run_hooks.py:260] loss = 1.0861053, step = 95000 (3.495 sec)
I0804 20:20:30.471377 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2491
I0804 20:20:30.472810 140200711067520 basic_session_run_hooks.py:260] loss = 1.1419286, step = 95100 (3.200 sec)
I0804 20:20:33.681154 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1549
I0804 20:20:33.682654 140200711067520 basic_session_run_hooks.py:260] loss = 1.0740303, step = 95200 (3.210 sec)
I0804 20:20:36.865331 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4053
I0804 20:20:36.866704 140200711067520 basic_session_run_hooks.py:260] loss = 1.1129447, step = 95300 (3.184 sec)
I0804 20:20:40.047587 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4244
I0804 20:20:40.049234 140200711067520 basic_session_run_hooks.py:260] loss = 1.1023118, step = 95400 (3.183 sec)
I0804 20:20:43.232610 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3968
I0804 20:20:43.234235 140200711067520 basic_session_run_hooks.py:260] loss = 1.1003395, step = 95500 (3.185 sec)
I0804 20:20:46.462559 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9604
I0804 20:20:46.463965 140200711067520 basic_session_run_hooks.py:260] loss = 1.0337193, step = 95600 (3.230 sec)
I0804 20:20:49.616197 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7091
I0804 20:20:49.617381 140200711067520 basic_session_run_hooks.py:260] loss = 1.0340748, step = 95700 (3.153 sec)
I0804 20:20:52.816078 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2513
I0804 20:20:52.817533 140200711067520 basic_session_run_hooks.py:260] loss = 1.0889533, step = 95800 (3.200 sec)
I0804 20:20:56.006012 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3487
I0804 20:20:56.007401 140200711067520 basic_session_run_hooks.py:260] loss = 1.0717409, step = 95900 (3.190 sec)
I0804 20:20:59.156440 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 96000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:20:59.436916 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:20:59.482445 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7649
I0804 20:20:59.483582 140200711067520 basic_session_run_hooks.py:260] loss = 1.1146753, step = 96000 (3.476 sec)
I0804 20:21:02.671618 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3562
I0804 20:21:02.673068 140200711067520 basic_session_run_hooks.py:260] loss = 1.0678753, step = 96100 (3.189 sec)
I0804 20:21:05.878586 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1823
I0804 20:21:05.879979 140200711067520 basic_session_run_hooks.py:260] loss = 1.0584409, step = 96200 (3.207 sec)
I0804 20:21:09.077377 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2617
I0804 20:21:09.078860 140200711067520 basic_session_run_hooks.py:260] loss = 1.0758984, step = 96300 (3.199 sec)
I0804 20:21:12.298738 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0426
I0804 20:21:12.300150 140200711067520 basic_session_run_hooks.py:260] loss = 1.0402173, step = 96400 (3.221 sec)
I0804 20:21:15.507598 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.164
I0804 20:21:15.509692 140200711067520 basic_session_run_hooks.py:260] loss = 1.0606382, step = 96500 (3.210 sec)
I0804 20:21:18.725810 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0731
I0804 20:21:18.727179 140200711067520 basic_session_run_hooks.py:260] loss = 1.0823917, step = 96600 (3.217 sec)
I0804 20:21:21.938351 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1278
I0804 20:21:21.939799 140200711067520 basic_session_run_hooks.py:260] loss = 1.1123354, step = 96700 (3.213 sec)
I0804 20:21:25.137526 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2584
I0804 20:21:25.139062 140200711067520 basic_session_run_hooks.py:260] loss = 1.1333286, step = 96800 (3.199 sec)
I0804 20:21:28.345924 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1678
I0804 20:21:28.347012 140200711067520 basic_session_run_hooks.py:260] loss = 1.0747583, step = 96900 (3.208 sec)
I0804 20:21:31.535109 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 97000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:21:31.829576 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:21:31.870442 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3727
I0804 20:21:31.871619 140200711067520 basic_session_run_hooks.py:260] loss = 1.028291, step = 97000 (3.525 sec)
I0804 20:21:35.067468 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2794
I0804 20:21:35.068673 140200711067520 basic_session_run_hooks.py:260] loss = 1.0462704, step = 97100 (3.197 sec)
I0804 20:21:38.252765 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.394
I0804 20:21:38.253810 140200711067520 basic_session_run_hooks.py:260] loss = 1.0077639, step = 97200 (3.185 sec)
I0804 20:21:41.447909 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2975
I0804 20:21:41.449353 140200711067520 basic_session_run_hooks.py:260] loss = 1.149525, step = 97300 (3.196 sec)
I0804 20:21:44.635690 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3698
I0804 20:21:44.637403 140200711067520 basic_session_run_hooks.py:260] loss = 1.0899152, step = 97400 (3.188 sec)
I0804 20:21:47.848274 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1275
I0804 20:21:47.849770 140200711067520 basic_session_run_hooks.py:260] loss = 1.0930177, step = 97500 (3.212 sec)
I0804 20:21:51.029913 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4303
I0804 20:21:51.031410 140200711067520 basic_session_run_hooks.py:260] loss = 1.0966009, step = 97600 (3.182 sec)
I0804 20:21:54.199858 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5463
I0804 20:21:54.201088 140200711067520 basic_session_run_hooks.py:260] loss = 1.10329, step = 97700 (3.170 sec)
I0804 20:21:57.363915 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6051
I0804 20:21:57.365469 140200711067520 basic_session_run_hooks.py:260] loss = 1.1222259, step = 97800 (3.164 sec)
I0804 20:22:00.561295 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2755
I0804 20:22:00.562723 140200711067520 basic_session_run_hooks.py:260] loss = 1.1449836, step = 97900 (3.197 sec)
I0804 20:22:03.741555 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 98000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:22:04.026864 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:22:04.068528 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5125
I0804 20:22:04.069832 140200711067520 basic_session_run_hooks.py:260] loss = 1.0777742, step = 98000 (3.507 sec)
I0804 20:22:07.260543 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3283
I0804 20:22:07.261737 140200711067520 basic_session_run_hooks.py:260] loss = 1.0800779, step = 98100 (3.192 sec)
I0804 20:22:10.434846 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5028
I0804 20:22:10.436213 140200711067520 basic_session_run_hooks.py:260] loss = 1.1446549, step = 98200 (3.174 sec)
I0804 20:22:13.646766 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1341
I0804 20:22:13.648009 140200711067520 basic_session_run_hooks.py:260] loss = 1.037677, step = 98300 (3.212 sec)
I0804 20:22:16.949925 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.274
I0804 20:22:16.951058 140200711067520 basic_session_run_hooks.py:260] loss = 1.1355284, step = 98400 (3.303 sec)
I0804 20:22:20.104941 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6957
I0804 20:22:20.106561 140200711067520 basic_session_run_hooks.py:260] loss = 1.0660923, step = 98500 (3.155 sec)
I0804 20:22:23.278074 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5147
I0804 20:22:23.279361 140200711067520 basic_session_run_hooks.py:260] loss = 1.1232358, step = 98600 (3.173 sec)
I0804 20:22:26.448565 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.541
I0804 20:22:26.450040 140200711067520 basic_session_run_hooks.py:260] loss = 1.0915672, step = 98700 (3.171 sec)
I0804 20:22:29.592884 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8031
I0804 20:22:29.594269 140200711067520 basic_session_run_hooks.py:260] loss = 1.0409738, step = 98800 (3.144 sec)
I0804 20:22:32.763869 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.536
I0804 20:22:32.765062 140200711067520 basic_session_run_hooks.py:260] loss = 1.1283987, step = 98900 (3.171 sec)
I0804 20:22:35.896963 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 99000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:22:36.189564 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:22:36.227825 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8685
I0804 20:22:36.228886 140200711067520 basic_session_run_hooks.py:260] loss = 1.0640594, step = 99000 (3.464 sec)
I0804 20:22:39.376411 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7607
I0804 20:22:39.377924 140200711067520 basic_session_run_hooks.py:260] loss = 1.0865963, step = 99100 (3.149 sec)
I0804 20:22:42.623199 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7994
I0804 20:22:42.625168 140200711067520 basic_session_run_hooks.py:260] loss = 1.1725997, step = 99200 (3.247 sec)
I0804 20:22:45.887670 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6328
I0804 20:22:45.889023 140200711067520 basic_session_run_hooks.py:260] loss = 1.0579427, step = 99300 (3.264 sec)
I0804 20:22:49.162648 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5346
I0804 20:22:49.163822 140200711067520 basic_session_run_hooks.py:260] loss = 1.146908, step = 99400 (3.275 sec)
I0804 20:22:52.431772 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5892
I0804 20:22:52.432857 140200711067520 basic_session_run_hooks.py:260] loss = 1.066684, step = 99500 (3.269 sec)
I0804 20:22:55.687992 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7105
I0804 20:22:55.689161 140200711067520 basic_session_run_hooks.py:260] loss = 1.1165813, step = 99600 (3.256 sec)
I0804 20:22:58.951769 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6392
I0804 20:22:58.953104 140200711067520 basic_session_run_hooks.py:260] loss = 1.1042123, step = 99700 (3.264 sec)
I0804 20:23:02.182284 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9548
I0804 20:23:02.183580 140200711067520 basic_session_run_hooks.py:260] loss = 1.1139923, step = 99800 (3.230 sec)
I0804 20:23:05.436939 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7255
I0804 20:23:05.438401 140200711067520 basic_session_run_hooks.py:260] loss = 1.0922121, step = 99900 (3.255 sec)
I0804 20:23:08.694411 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 100000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:23:08.988560 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:23:09.022504 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 27.8894
I0804 20:23:09.023653 140200711067520 basic_session_run_hooks.py:260] loss = 1.098027, step = 100000 (3.585 sec)
I0804 20:23:12.239777 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0822
I0804 20:23:12.241082 140200711067520 basic_session_run_hooks.py:260] loss = 1.1307608, step = 100100 (3.217 sec)
I0804 20:23:15.468060 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9763
I0804 20:23:15.469581 140200711067520 basic_session_run_hooks.py:260] loss = 1.0565836, step = 100200 (3.228 sec)
I0804 20:23:18.680850 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1254
I0804 20:23:18.681928 140200711067520 basic_session_run_hooks.py:260] loss = 1.1106963, step = 100300 (3.212 sec)
I0804 20:23:21.880981 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2492
I0804 20:23:21.882663 140200711067520 basic_session_run_hooks.py:260] loss = 1.0795224, step = 100400 (3.201 sec)
I0804 20:23:25.116972 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9021
I0804 20:23:25.118169 140200711067520 basic_session_run_hooks.py:260] loss = 1.1646417, step = 100500 (3.236 sec)
I0804 20:23:28.319988 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2205
I0804 20:23:28.321534 140200711067520 basic_session_run_hooks.py:260] loss = 1.0931497, step = 100600 (3.203 sec)
I0804 20:23:31.531328 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1396
I0804 20:23:31.532644 140200711067520 basic_session_run_hooks.py:260] loss = 1.0481747, step = 100700 (3.211 sec)
I0804 20:23:34.789581 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6916
I0804 20:23:34.791026 140200711067520 basic_session_run_hooks.py:260] loss = 1.1893134, step = 100800 (3.258 sec)
I0804 20:23:38.009733 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0541
I0804 20:23:38.011065 140200711067520 basic_session_run_hooks.py:260] loss = 1.0680556, step = 100900 (3.220 sec)
I0804 20:23:41.220122 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 101000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:23:41.516078 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:23:41.558303 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.1803
I0804 20:23:41.559598 140200711067520 basic_session_run_hooks.py:260] loss = 1.1250961, step = 101000 (3.549 sec)
I0804 20:23:44.790018 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9435
I0804 20:23:44.791159 140200711067520 basic_session_run_hooks.py:260] loss = 1.0436763, step = 101100 (3.232 sec)
I0804 20:23:48.026437 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8985
I0804 20:23:48.027758 140200711067520 basic_session_run_hooks.py:260] loss = 1.0946741, step = 101200 (3.237 sec)
I0804 20:23:51.258946 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9357
I0804 20:23:51.260435 140200711067520 basic_session_run_hooks.py:260] loss = 1.1190228, step = 101300 (3.233 sec)
I0804 20:23:54.462238 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2179
I0804 20:23:54.463883 140200711067520 basic_session_run_hooks.py:260] loss = 1.0754343, step = 101400 (3.203 sec)
I0804 20:23:57.678726 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0897
I0804 20:23:57.680279 140200711067520 basic_session_run_hooks.py:260] loss = 1.10015, step = 101500 (3.216 sec)
I0804 20:24:00.914764 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9019
I0804 20:24:00.916040 140200711067520 basic_session_run_hooks.py:260] loss = 1.1015984, step = 101600 (3.236 sec)
I0804 20:24:04.136770 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0367
I0804 20:24:04.138068 140200711067520 basic_session_run_hooks.py:260] loss = 1.087062, step = 101700 (3.222 sec)
I0804 20:24:07.339203 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2262
I0804 20:24:07.340690 140200711067520 basic_session_run_hooks.py:260] loss = 1.124545, step = 101800 (3.203 sec)
I0804 20:24:10.573803 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.916
I0804 20:24:10.575475 140200711067520 basic_session_run_hooks.py:260] loss = 1.1478754, step = 101900 (3.235 sec)
I0804 20:24:13.748300 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 102000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:24:14.045460 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:24:14.085644 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4748
I0804 20:24:14.086787 140200711067520 basic_session_run_hooks.py:260] loss = 1.0704353, step = 102000 (3.511 sec)
I0804 20:24:17.322763 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8919
I0804 20:24:17.323987 140200711067520 basic_session_run_hooks.py:260] loss = 1.1242783, step = 102100 (3.237 sec)
I0804 20:24:20.552834 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.959
I0804 20:24:20.554009 140200711067520 basic_session_run_hooks.py:260] loss = 1.0622567, step = 102200 (3.230 sec)
I0804 20:24:23.823230 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5773
I0804 20:24:23.824632 140200711067520 basic_session_run_hooks.py:260] loss = 1.0354576, step = 102300 (3.271 sec)
I0804 20:24:27.071539 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7855
I0804 20:24:27.072861 140200711067520 basic_session_run_hooks.py:260] loss = 1.1544284, step = 102400 (3.248 sec)
I0804 20:24:30.279660 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1711
I0804 20:24:30.281397 140200711067520 basic_session_run_hooks.py:260] loss = 1.1284912, step = 102500 (3.209 sec)
I0804 20:24:33.535478 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.714
I0804 20:24:33.536845 140200711067520 basic_session_run_hooks.py:260] loss = 1.0946177, step = 102600 (3.255 sec)
I0804 20:24:36.778773 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8326
I0804 20:24:36.780050 140200711067520 basic_session_run_hooks.py:260] loss = 1.1095964, step = 102700 (3.243 sec)
I0804 20:24:39.994294 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0994
I0804 20:24:39.995901 140200711067520 basic_session_run_hooks.py:260] loss = 1.1518661, step = 102800 (3.216 sec)
I0804 20:24:43.217823 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0218
I0804 20:24:43.219237 140200711067520 basic_session_run_hooks.py:260] loss = 1.0669605, step = 102900 (3.223 sec)
I0804 20:24:46.417732 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 103000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:24:46.728080 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:24:46.771268 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.1414
I0804 20:24:46.772271 140200711067520 basic_session_run_hooks.py:260] loss = 0.9929067, step = 103000 (3.553 sec)
I0804 20:24:50.051518 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.486
I0804 20:24:50.052821 140200711067520 basic_session_run_hooks.py:260] loss = 1.0869904, step = 103100 (3.281 sec)
I0804 20:24:53.286131 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9153
I0804 20:24:53.287353 140200711067520 basic_session_run_hooks.py:260] loss = 1.0957005, step = 103200 (3.235 sec)
I0804 20:24:56.531694 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8113
I0804 20:24:56.532891 140200711067520 basic_session_run_hooks.py:260] loss = 1.1107419, step = 103300 (3.246 sec)
I0804 20:24:59.753333 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.04
I0804 20:24:59.754709 140200711067520 basic_session_run_hooks.py:260] loss = 1.0138078, step = 103400 (3.222 sec)
I0804 20:25:02.963162 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1544
I0804 20:25:02.964850 140200711067520 basic_session_run_hooks.py:260] loss = 1.0364459, step = 103500 (3.210 sec)
I0804 20:25:06.163626 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2458
I0804 20:25:06.165132 140200711067520 basic_session_run_hooks.py:260] loss = 1.0927892, step = 103600 (3.200 sec)
I0804 20:25:09.356651 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3179
I0804 20:25:09.358072 140200711067520 basic_session_run_hooks.py:260] loss = 1.147671, step = 103700 (3.193 sec)
I0804 20:25:12.557663 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2401
I0804 20:25:12.559181 140200711067520 basic_session_run_hooks.py:260] loss = 1.026202, step = 103800 (3.201 sec)
I0804 20:25:15.794735 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8922
I0804 20:25:15.795804 140200711067520 basic_session_run_hooks.py:260] loss = 1.0305007, step = 103900 (3.237 sec)
I0804 20:25:18.981331 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 104000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:25:19.270987 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:25:19.320990 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3585
I0804 20:25:19.322236 140200711067520 basic_session_run_hooks.py:260] loss = 1.1267533, step = 104000 (3.526 sec)
I0804 20:25:22.538820 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0776
I0804 20:25:22.540156 140200711067520 basic_session_run_hooks.py:260] loss = 1.061093, step = 104100 (3.218 sec)
I0804 20:25:25.738174 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2557
I0804 20:25:25.739686 140200711067520 basic_session_run_hooks.py:260] loss = 1.0489335, step = 104200 (3.200 sec)
I0804 20:25:28.945036 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1832
I0804 20:25:28.946512 140200711067520 basic_session_run_hooks.py:260] loss = 1.0465553, step = 104300 (3.207 sec)
I0804 20:25:32.160701 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0978
I0804 20:25:32.161867 140200711067520 basic_session_run_hooks.py:260] loss = 0.98387593, step = 104400 (3.215 sec)
I0804 20:25:35.386440 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0007
I0804 20:25:35.387753 140200711067520 basic_session_run_hooks.py:260] loss = 1.0575451, step = 104500 (3.226 sec)
I0804 20:25:38.584284 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2709
I0804 20:25:38.585756 140200711067520 basic_session_run_hooks.py:260] loss = 1.092471, step = 104600 (3.198 sec)
I0804 20:25:41.831372 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7968
I0804 20:25:41.832922 140200711067520 basic_session_run_hooks.py:260] loss = 1.1135404, step = 104700 (3.247 sec)
I0804 20:25:45.030400 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2596
I0804 20:25:45.031833 140200711067520 basic_session_run_hooks.py:260] loss = 1.1385401, step = 104800 (3.199 sec)
I0804 20:25:48.225395 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2989
I0804 20:25:48.226871 140200711067520 basic_session_run_hooks.py:260] loss = 1.0430063, step = 104900 (3.195 sec)
I0804 20:25:51.389029 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 105000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:25:51.673414 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:25:51.708732 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7079
I0804 20:25:51.709840 140200711067520 basic_session_run_hooks.py:260] loss = 1.0239401, step = 105000 (3.483 sec)
I0804 20:25:54.923295 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1087
I0804 20:25:54.924557 140200711067520 basic_session_run_hooks.py:260] loss = 1.1158227, step = 105100 (3.215 sec)
I0804 20:25:58.112351 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3573
I0804 20:25:58.114356 140200711067520 basic_session_run_hooks.py:260] loss = 1.1230214, step = 105200 (3.190 sec)
I0804 20:26:01.301640 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3548
I0804 20:26:01.303061 140200711067520 basic_session_run_hooks.py:260] loss = 1.1754649, step = 105300 (3.189 sec)
I0804 20:26:04.496090 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3044
I0804 20:26:04.497765 140200711067520 basic_session_run_hooks.py:260] loss = 1.1392729, step = 105400 (3.195 sec)
I0804 20:26:07.715156 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.065
I0804 20:26:07.716514 140200711067520 basic_session_run_hooks.py:260] loss = 1.1132761, step = 105500 (3.219 sec)
I0804 20:26:10.881641 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5806
I0804 20:26:10.883036 140200711067520 basic_session_run_hooks.py:260] loss = 1.0413076, step = 105600 (3.167 sec)
I0804 20:26:14.035775 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7043
I0804 20:26:14.037056 140200711067520 basic_session_run_hooks.py:260] loss = 1.0542986, step = 105700 (3.154 sec)
I0804 20:26:17.187261 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7312
I0804 20:26:17.188636 140200711067520 basic_session_run_hooks.py:260] loss = 1.0093136, step = 105800 (3.152 sec)
I0804 20:26:20.344984 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6683
I0804 20:26:20.346391 140200711067520 basic_session_run_hooks.py:260] loss = 1.1207507, step = 105900 (3.158 sec)
I0804 20:26:23.491306 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 106000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:26:23.796971 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 20:26:23.798483 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 20:26:23.947068 140200711067520 estimator.py:1145] Calling model_fn.
I0804 20:26:23.947998 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 20:26:23.948390 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 20:26:23.948496 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 20:26:23.948581 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 20:26:23.948648 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 20:26:23.948730 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 20:26:23.948802 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 20:26:24.035287 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 20:26:24.095023 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 20:26:24.233679 140200711067520 t2t_model.py:2172] Building model body
I0804 20:26:25.192487 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 20:26:25.907534 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 20:26:25.925642 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T20:26:25Z
I0804 20:26:26.091324 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 20:26:26.092034: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:26:26.092438: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 20:26:26.092531: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 20:26:26.092553: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 20:26:26.092571: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 20:26:26.092591: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 20:26:26.092614: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 20:26:26.092633: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 20:26:26.092653: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 20:26:26.092758: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:26:26.093164: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:26:26.093498: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 20:26:26.093541: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 20:26:26.093555: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 20:26:26.093565: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 20:26:26.093846: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:26:26.094241: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:26:26.094580: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 20:26:26.096177 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-106000
I0804 20:26:26.300190 140200711067520 session_manager.py:500] Running local_init_op.
I0804 20:26:26.344919 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 20:26:32.389020 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 20:26:37.706038 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 20:26:43.037734 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 20:26:48.404938 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 20:26:53.802503 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 20:26:59.184478 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 20:27:04.517393 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 20:27:09.879665 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 20:27:15.198533 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 20:27:20.039729 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-20:27:20
I0804 20:27:20.039967 140200711067520 estimator.py:2039] Saving dict for global step 106000: global_step = 106000, loss = 1.1833342, metrics-paper_generation_problem/targets/accuracy = 0.67203283, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8833171, metrics-paper_generation_problem/targets/approx_bleu_score = 0.48583966, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1833702, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5805998, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.69318026
I0804 20:27:20.040413 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 106000: experiment/transformer/transformer_small/output/model.ckpt-106000
I0804 20:27:20.093387 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.67368
I0804 20:27:20.094609 140200711067520 basic_session_run_hooks.py:260] loss = 1.0309321, step = 106000 (59.748 sec)
I0804 20:27:23.319628 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9962
I0804 20:27:23.321094 140200711067520 basic_session_run_hooks.py:260] loss = 1.0246035, step = 106100 (3.226 sec)
I0804 20:27:26.462165 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8213
I0804 20:27:26.463746 140200711067520 basic_session_run_hooks.py:260] loss = 1.1383594, step = 106200 (3.143 sec)
I0804 20:27:29.680401 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.073
I0804 20:27:29.681718 140200711067520 basic_session_run_hooks.py:260] loss = 1.0154268, step = 106300 (3.218 sec)
I0804 20:27:32.924340 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8267
I0804 20:27:32.925641 140200711067520 basic_session_run_hooks.py:260] loss = 1.1065768, step = 106400 (3.244 sec)
I0804 20:27:36.170281 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8076
I0804 20:27:36.171590 140200711067520 basic_session_run_hooks.py:260] loss = 1.1365702, step = 106500 (3.246 sec)
I0804 20:27:39.453778 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4554
I0804 20:27:39.455477 140200711067520 basic_session_run_hooks.py:260] loss = 1.0785077, step = 106600 (3.284 sec)
I0804 20:27:42.731065 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5131
I0804 20:27:42.732582 140200711067520 basic_session_run_hooks.py:260] loss = 1.0887903, step = 106700 (3.277 sec)
I0804 20:27:45.991267 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6729
I0804 20:27:45.992885 140200711067520 basic_session_run_hooks.py:260] loss = 1.1014143, step = 106800 (3.260 sec)
I0804 20:27:49.253982 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6493
I0804 20:27:49.255441 140200711067520 basic_session_run_hooks.py:260] loss = 1.0769331, step = 106900 (3.263 sec)
I0804 20:27:52.482513 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 107000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:27:52.780040 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:27:52.819368 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.0472
I0804 20:27:52.820382 140200711067520 basic_session_run_hooks.py:260] loss = 1.0805478, step = 107000 (3.565 sec)
I0804 20:27:56.027706 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.169
I0804 20:27:56.029132 140200711067520 basic_session_run_hooks.py:260] loss = 1.1153398, step = 107100 (3.209 sec)
I0804 20:27:59.223130 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2947
I0804 20:27:59.224213 140200711067520 basic_session_run_hooks.py:260] loss = 1.0742234, step = 107200 (3.195 sec)
I0804 20:28:02.440212 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0841
I0804 20:28:02.441604 140200711067520 basic_session_run_hooks.py:260] loss = 1.0667528, step = 107300 (3.217 sec)
I0804 20:28:05.630309 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.347
I0804 20:28:05.631551 140200711067520 basic_session_run_hooks.py:260] loss = 1.0479318, step = 107400 (3.190 sec)
I0804 20:28:08.831170 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2417
I0804 20:28:08.832666 140200711067520 basic_session_run_hooks.py:260] loss = 1.0750887, step = 107500 (3.201 sec)
I0804 20:28:12.029818 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2632
I0804 20:28:12.031109 140200711067520 basic_session_run_hooks.py:260] loss = 1.0754836, step = 107600 (3.198 sec)
I0804 20:28:15.235270 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1971
I0804 20:28:15.236652 140200711067520 basic_session_run_hooks.py:260] loss = 1.0385772, step = 107700 (3.206 sec)
I0804 20:28:18.440628 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1975
I0804 20:28:18.441866 140200711067520 basic_session_run_hooks.py:260] loss = 1.177805, step = 107800 (3.205 sec)
I0804 20:28:21.601528 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6374
I0804 20:28:21.602956 140200711067520 basic_session_run_hooks.py:260] loss = 1.0668449, step = 107900 (3.161 sec)
I0804 20:28:24.703793 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 108000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:28:24.996773 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:28:25.039455 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0865
I0804 20:28:25.040514 140200711067520 basic_session_run_hooks.py:260] loss = 1.1579063, step = 108000 (3.438 sec)
I0804 20:28:28.197591 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6645
I0804 20:28:28.198883 140200711067520 basic_session_run_hooks.py:260] loss = 1.0632963, step = 108100 (3.158 sec)
I0804 20:28:31.271931 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.5272
I0804 20:28:31.273382 140200711067520 basic_session_run_hooks.py:260] loss = 1.0515679, step = 108200 (3.075 sec)
I0804 20:28:34.477329 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1973
I0804 20:28:34.478630 140200711067520 basic_session_run_hooks.py:260] loss = 1.1127657, step = 108300 (3.205 sec)
I0804 20:28:37.627155 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.748
I0804 20:28:37.628586 140200711067520 basic_session_run_hooks.py:260] loss = 1.0931996, step = 108400 (3.150 sec)
I0804 20:28:40.767880 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8398
I0804 20:28:40.769345 140200711067520 basic_session_run_hooks.py:260] loss = 1.1152681, step = 108500 (3.141 sec)
I0804 20:28:43.941808 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5066
I0804 20:28:43.943320 140200711067520 basic_session_run_hooks.py:260] loss = 1.1431539, step = 108600 (3.174 sec)
I0804 20:28:47.121398 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4507
I0804 20:28:47.122625 140200711067520 basic_session_run_hooks.py:260] loss = 1.1020706, step = 108700 (3.179 sec)
I0804 20:28:50.298640 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4737
I0804 20:28:50.300061 140200711067520 basic_session_run_hooks.py:260] loss = 1.0374557, step = 108800 (3.177 sec)
I0804 20:28:53.482983 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4037
I0804 20:28:53.484493 140200711067520 basic_session_run_hooks.py:260] loss = 1.1357403, step = 108900 (3.184 sec)
I0804 20:28:56.611725 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 109000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:28:56.902654 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:28:56.941261 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.916
I0804 20:28:56.942259 140200711067520 basic_session_run_hooks.py:260] loss = 1.063018, step = 109000 (3.458 sec)
I0804 20:29:00.153337 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1328
I0804 20:29:00.154807 140200711067520 basic_session_run_hooks.py:260] loss = 1.0860237, step = 109100 (3.213 sec)
I0804 20:29:03.327606 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5032
I0804 20:29:03.329061 140200711067520 basic_session_run_hooks.py:260] loss = 1.0488541, step = 109200 (3.174 sec)
I0804 20:29:06.512576 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3977
I0804 20:29:06.513931 140200711067520 basic_session_run_hooks.py:260] loss = 1.0588373, step = 109300 (3.185 sec)
I0804 20:29:09.686586 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5059
I0804 20:29:09.687920 140200711067520 basic_session_run_hooks.py:260] loss = 1.0659906, step = 109400 (3.174 sec)
I0804 20:29:12.890400 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2127
I0804 20:29:12.891985 140200711067520 basic_session_run_hooks.py:260] loss = 1.0380012, step = 109500 (3.204 sec)
I0804 20:29:16.073148 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4193
I0804 20:29:16.074623 140200711067520 basic_session_run_hooks.py:260] loss = 1.0725542, step = 109600 (3.183 sec)
I0804 20:29:19.256925 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4095
I0804 20:29:19.258295 140200711067520 basic_session_run_hooks.py:260] loss = 1.1131474, step = 109700 (3.184 sec)
I0804 20:29:22.454790 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2707
I0804 20:29:22.456103 140200711067520 basic_session_run_hooks.py:260] loss = 1.0445925, step = 109800 (3.198 sec)
I0804 20:29:25.710470 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7155
I0804 20:29:25.711930 140200711067520 basic_session_run_hooks.py:260] loss = 1.0973185, step = 109900 (3.256 sec)
I0804 20:29:28.875710 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 110000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:29:29.179175 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:29:29.220930 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4861
I0804 20:29:29.222139 140200711067520 basic_session_run_hooks.py:260] loss = 1.0913119, step = 110000 (3.510 sec)
I0804 20:29:32.402596 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4305
I0804 20:29:32.403958 140200711067520 basic_session_run_hooks.py:260] loss = 1.039239, step = 110100 (3.182 sec)
I0804 20:29:35.592268 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3511
I0804 20:29:35.593777 140200711067520 basic_session_run_hooks.py:260] loss = 1.0646951, step = 110200 (3.190 sec)
I0804 20:29:38.794120 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.232
I0804 20:29:38.795308 140200711067520 basic_session_run_hooks.py:260] loss = 1.0816709, step = 110300 (3.202 sec)
I0804 20:29:41.961655 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5701
I0804 20:29:41.963022 140200711067520 basic_session_run_hooks.py:260] loss = 1.1507891, step = 110400 (3.168 sec)
I0804 20:29:45.150182 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3625
I0804 20:29:45.151547 140200711067520 basic_session_run_hooks.py:260] loss = 1.1051593, step = 110500 (3.189 sec)
I0804 20:29:48.346933 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2818
I0804 20:29:48.348115 140200711067520 basic_session_run_hooks.py:260] loss = 1.0825691, step = 110600 (3.197 sec)
I0804 20:29:51.575091 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9773
I0804 20:29:51.576464 140200711067520 basic_session_run_hooks.py:260] loss = 1.0235887, step = 110700 (3.228 sec)
I0804 20:29:54.723762 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7595
I0804 20:29:54.725304 140200711067520 basic_session_run_hooks.py:260] loss = 1.0806336, step = 110800 (3.149 sec)
I0804 20:29:57.897380 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5098
I0804 20:29:57.898917 140200711067520 basic_session_run_hooks.py:260] loss = 1.0535134, step = 110900 (3.174 sec)
I0804 20:30:01.084610 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 111000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:30:01.378505 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:30:01.417237 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4102
I0804 20:30:01.418840 140200711067520 basic_session_run_hooks.py:260] loss = 1.0503011, step = 111000 (3.520 sec)
I0804 20:30:04.623139 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1926
I0804 20:30:04.624241 140200711067520 basic_session_run_hooks.py:260] loss = 1.0773273, step = 111100 (3.205 sec)
I0804 20:30:07.827342 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.209
I0804 20:30:07.828626 140200711067520 basic_session_run_hooks.py:260] loss = 1.1401334, step = 111200 (3.204 sec)
I0804 20:30:11.009114 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.429
I0804 20:30:11.010555 140200711067520 basic_session_run_hooks.py:260] loss = 1.0352199, step = 111300 (3.182 sec)
I0804 20:30:14.202826 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3115
I0804 20:30:14.204291 140200711067520 basic_session_run_hooks.py:260] loss = 0.998941, step = 111400 (3.194 sec)
I0804 20:30:17.404190 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2367
I0804 20:30:17.405555 140200711067520 basic_session_run_hooks.py:260] loss = 1.0080043, step = 111500 (3.201 sec)
I0804 20:30:20.605464 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2379
I0804 20:30:20.606723 140200711067520 basic_session_run_hooks.py:260] loss = 1.0661896, step = 111600 (3.201 sec)
I0804 20:30:23.821305 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0958
I0804 20:30:23.822830 140200711067520 basic_session_run_hooks.py:260] loss = 1.1467514, step = 111700 (3.216 sec)
I0804 20:30:27.035147 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1154
I0804 20:30:27.036196 140200711067520 basic_session_run_hooks.py:260] loss = 1.0682925, step = 111800 (3.213 sec)
I0804 20:30:30.234702 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2544
I0804 20:30:30.235924 140200711067520 basic_session_run_hooks.py:260] loss = 1.0794193, step = 111900 (3.200 sec)
I0804 20:30:33.397903 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 112000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:30:33.686684 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:30:33.735900 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5615
I0804 20:30:33.737054 140200711067520 basic_session_run_hooks.py:260] loss = 1.0995529, step = 112000 (3.501 sec)
I0804 20:30:36.914682 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4587
I0804 20:30:36.915817 140200711067520 basic_session_run_hooks.py:260] loss = 1.1083369, step = 112100 (3.179 sec)
I0804 20:30:40.134292 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0597
I0804 20:30:40.135789 140200711067520 basic_session_run_hooks.py:260] loss = 1.1406009, step = 112200 (3.220 sec)
I0804 20:30:43.300756 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5809
I0804 20:30:43.301979 140200711067520 basic_session_run_hooks.py:260] loss = 1.1187463, step = 112300 (3.166 sec)
I0804 20:30:46.445370 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8005
I0804 20:30:46.446935 140200711067520 basic_session_run_hooks.py:260] loss = 1.1131047, step = 112400 (3.145 sec)
I0804 20:30:49.623203 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.468
I0804 20:30:49.624580 140200711067520 basic_session_run_hooks.py:260] loss = 1.0607682, step = 112500 (3.178 sec)
I0804 20:30:52.812485 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3551
I0804 20:30:52.814054 140200711067520 basic_session_run_hooks.py:260] loss = 1.1723627, step = 112600 (3.189 sec)
I0804 20:30:55.979943 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.571
I0804 20:30:55.981262 140200711067520 basic_session_run_hooks.py:260] loss = 1.0990022, step = 112700 (3.167 sec)
I0804 20:30:59.148627 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5588
I0804 20:30:59.150013 140200711067520 basic_session_run_hooks.py:260] loss = 1.0189258, step = 112800 (3.169 sec)
I0804 20:31:02.325227 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4801
I0804 20:31:02.326332 140200711067520 basic_session_run_hooks.py:260] loss = 1.0538574, step = 112900 (3.176 sec)
I0804 20:31:05.491662 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 113000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:31:05.787626 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:31:05.832723 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5102
I0804 20:31:05.833755 140200711067520 basic_session_run_hooks.py:260] loss = 1.0542742, step = 113000 (3.507 sec)
I0804 20:31:09.012171 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4522
I0804 20:31:09.013639 140200711067520 basic_session_run_hooks.py:260] loss = 1.1364133, step = 113100 (3.180 sec)
I0804 20:31:12.214201 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2302
I0804 20:31:12.215450 140200711067520 basic_session_run_hooks.py:260] loss = 1.1219925, step = 113200 (3.202 sec)
I0804 20:31:15.390163 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4866
I0804 20:31:15.391524 140200711067520 basic_session_run_hooks.py:260] loss = 1.0524733, step = 113300 (3.176 sec)
I0804 20:31:18.544685 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7006
I0804 20:31:18.546047 140200711067520 basic_session_run_hooks.py:260] loss = 1.1581272, step = 113400 (3.155 sec)
I0804 20:31:21.721614 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.477
I0804 20:31:21.723077 140200711067520 basic_session_run_hooks.py:260] loss = 1.1048352, step = 113500 (3.177 sec)
I0804 20:31:24.923273 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2336
I0804 20:31:24.924804 140200711067520 basic_session_run_hooks.py:260] loss = 1.1150696, step = 113600 (3.202 sec)
I0804 20:31:28.098200 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4968
I0804 20:31:28.099324 140200711067520 basic_session_run_hooks.py:260] loss = 1.0508865, step = 113700 (3.175 sec)
I0804 20:31:31.315093 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0859
I0804 20:31:31.316552 140200711067520 basic_session_run_hooks.py:260] loss = 1.0133332, step = 113800 (3.217 sec)
I0804 20:31:34.511516 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2851
I0804 20:31:34.512986 140200711067520 basic_session_run_hooks.py:260] loss = 1.1207043, step = 113900 (3.196 sec)
I0804 20:31:37.692487 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 114000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:31:37.975999 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:31:38.022162 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4845
I0804 20:31:38.023350 140200711067520 basic_session_run_hooks.py:260] loss = 1.1608677, step = 114000 (3.510 sec)
I0804 20:31:41.240622 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0709
I0804 20:31:41.241999 140200711067520 basic_session_run_hooks.py:260] loss = 1.1214155, step = 114100 (3.219 sec)
I0804 20:31:44.450721 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1517
I0804 20:31:44.452065 140200711067520 basic_session_run_hooks.py:260] loss = 0.9920795, step = 114200 (3.210 sec)
I0804 20:31:47.660040 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1592
I0804 20:31:47.661281 140200711067520 basic_session_run_hooks.py:260] loss = 1.1276854, step = 114300 (3.209 sec)
I0804 20:31:50.878102 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0747
I0804 20:31:50.879590 140200711067520 basic_session_run_hooks.py:260] loss = 1.1191103, step = 114400 (3.218 sec)
I0804 20:31:54.090967 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1248
I0804 20:31:54.092233 140200711067520 basic_session_run_hooks.py:260] loss = 1.1196064, step = 114500 (3.213 sec)
I0804 20:31:57.319918 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9699
I0804 20:31:57.321185 140200711067520 basic_session_run_hooks.py:260] loss = 1.0790743, step = 114600 (3.229 sec)
I0804 20:32:00.457785 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8687
I0804 20:32:00.459105 140200711067520 basic_session_run_hooks.py:260] loss = 1.0893673, step = 114700 (3.138 sec)
I0804 20:32:03.605956 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7645
I0804 20:32:03.607202 140200711067520 basic_session_run_hooks.py:260] loss = 1.028806, step = 114800 (3.148 sec)
I0804 20:32:06.750839 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7977
I0804 20:32:06.752468 140200711067520 basic_session_run_hooks.py:260] loss = 1.1005218, step = 114900 (3.145 sec)
I0804 20:32:09.878168 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 115000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:32:10.170964 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:32:10.211572 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8954
I0804 20:32:10.212679 140200711067520 basic_session_run_hooks.py:260] loss = 0.987777, step = 115000 (3.460 sec)
I0804 20:32:13.362787 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.734
I0804 20:32:13.363906 140200711067520 basic_session_run_hooks.py:260] loss = 1.1296916, step = 115100 (3.151 sec)
I0804 20:32:16.529244 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5811
I0804 20:32:16.530510 140200711067520 basic_session_run_hooks.py:260] loss = 1.1546773, step = 115200 (3.167 sec)
I0804 20:32:19.701374 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5247
I0804 20:32:19.702742 140200711067520 basic_session_run_hooks.py:260] loss = 1.1237406, step = 115300 (3.172 sec)
I0804 20:32:22.900715 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2563
I0804 20:32:22.902156 140200711067520 basic_session_run_hooks.py:260] loss = 1.0841187, step = 115400 (3.199 sec)
I0804 20:32:26.076193 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4915
I0804 20:32:26.077500 140200711067520 basic_session_run_hooks.py:260] loss = 1.0101086, step = 115500 (3.175 sec)
I0804 20:32:29.276546 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2466
I0804 20:32:29.277970 140200711067520 basic_session_run_hooks.py:260] loss = 1.1346307, step = 115600 (3.200 sec)
I0804 20:32:32.472250 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2918
I0804 20:32:32.473577 140200711067520 basic_session_run_hooks.py:260] loss = 1.0481305, step = 115700 (3.196 sec)
I0804 20:32:35.661865 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3518
I0804 20:32:35.663142 140200711067520 basic_session_run_hooks.py:260] loss = 1.123434, step = 115800 (3.190 sec)
I0804 20:32:38.854172 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3253
I0804 20:32:38.855753 140200711067520 basic_session_run_hooks.py:260] loss = 1.1112989, step = 115900 (3.193 sec)
I0804 20:32:42.033877 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 116000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:32:42.320863 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:32:42.363997 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4913
I0804 20:32:42.365170 140200711067520 basic_session_run_hooks.py:260] loss = 1.1336573, step = 116000 (3.509 sec)
I0804 20:32:45.564989 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2406
I0804 20:32:45.566522 140200711067520 basic_session_run_hooks.py:260] loss = 1.0889307, step = 116100 (3.201 sec)
I0804 20:32:48.762285 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2763
I0804 20:32:48.763716 140200711067520 basic_session_run_hooks.py:260] loss = 1.0538504, step = 116200 (3.197 sec)
I0804 20:32:51.969434 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1805
I0804 20:32:51.970655 140200711067520 basic_session_run_hooks.py:260] loss = 1.0937747, step = 116300 (3.207 sec)
I0804 20:32:55.193207 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0194
I0804 20:32:55.194653 140200711067520 basic_session_run_hooks.py:260] loss = 1.0741394, step = 116400 (3.224 sec)
I0804 20:32:58.393649 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2456
I0804 20:32:58.395086 140200711067520 basic_session_run_hooks.py:260] loss = 1.0553746, step = 116500 (3.200 sec)
I0804 20:33:01.566811 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5143
I0804 20:33:01.568122 140200711067520 basic_session_run_hooks.py:260] loss = 1.0738312, step = 116600 (3.173 sec)
I0804 20:33:04.776273 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1581
I0804 20:33:04.777768 140200711067520 basic_session_run_hooks.py:260] loss = 1.115229, step = 116700 (3.210 sec)
I0804 20:33:07.994368 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0742
I0804 20:33:07.995940 140200711067520 basic_session_run_hooks.py:260] loss = 1.1424474, step = 116800 (3.218 sec)
I0804 20:33:11.194064 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2529
I0804 20:33:11.195373 140200711067520 basic_session_run_hooks.py:260] loss = 1.0305227, step = 116900 (3.199 sec)
I0804 20:33:14.338582 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 117000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:33:14.633831 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:33:14.672519 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7483
I0804 20:33:14.673516 140200711067520 basic_session_run_hooks.py:260] loss = 1.155822, step = 117000 (3.478 sec)
I0804 20:33:17.844954 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5217
I0804 20:33:17.846356 140200711067520 basic_session_run_hooks.py:260] loss = 1.0329493, step = 117100 (3.173 sec)
I0804 20:33:21.033553 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3618
I0804 20:33:21.034944 140200711067520 basic_session_run_hooks.py:260] loss = 1.0604116, step = 117200 (3.189 sec)
I0804 20:33:24.222589 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3576
I0804 20:33:24.224010 140200711067520 basic_session_run_hooks.py:260] loss = 1.0909264, step = 117300 (3.189 sec)
I0804 20:33:27.424215 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.234
I0804 20:33:27.425710 140200711067520 basic_session_run_hooks.py:260] loss = 1.0424238, step = 117400 (3.202 sec)
I0804 20:33:30.617444 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3164
I0804 20:33:30.618828 140200711067520 basic_session_run_hooks.py:260] loss = 1.0489106, step = 117500 (3.193 sec)
I0804 20:33:33.808168 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3407
I0804 20:33:33.809404 140200711067520 basic_session_run_hooks.py:260] loss = 1.0333016, step = 117600 (3.191 sec)
I0804 20:33:36.989456 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.434
I0804 20:33:36.991052 140200711067520 basic_session_run_hooks.py:260] loss = 1.1181022, step = 117700 (3.182 sec)
I0804 20:33:40.161496 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5255
I0804 20:33:40.162927 140200711067520 basic_session_run_hooks.py:260] loss = 1.1541423, step = 117800 (3.172 sec)
I0804 20:33:43.306283 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7985
I0804 20:33:43.307687 140200711067520 basic_session_run_hooks.py:260] loss = 1.0339297, step = 117900 (3.145 sec)
I0804 20:33:46.420527 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 118000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:33:46.721226 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:33:46.763412 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.9254
I0804 20:33:46.764494 140200711067520 basic_session_run_hooks.py:260] loss = 1.1155432, step = 118000 (3.457 sec)
I0804 20:33:49.924616 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6338
I0804 20:33:49.925923 140200711067520 basic_session_run_hooks.py:260] loss = 1.1288897, step = 118100 (3.161 sec)
I0804 20:33:53.218951 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.3552
I0804 20:33:53.220235 140200711067520 basic_session_run_hooks.py:260] loss = 1.1736562, step = 118200 (3.294 sec)
I0804 20:33:56.380143 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6336
I0804 20:33:56.381698 140200711067520 basic_session_run_hooks.py:260] loss = 1.0984596, step = 118300 (3.161 sec)
I0804 20:33:59.516609 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8829
I0804 20:33:59.517890 140200711067520 basic_session_run_hooks.py:260] loss = 1.076051, step = 118400 (3.136 sec)
I0804 20:34:02.683075 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5811
I0804 20:34:02.684697 140200711067520 basic_session_run_hooks.py:260] loss = 1.1042825, step = 118500 (3.167 sec)
I0804 20:34:05.837449 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7022
I0804 20:34:05.838648 140200711067520 basic_session_run_hooks.py:260] loss = 1.1262591, step = 118600 (3.154 sec)
I0804 20:34:08.986519 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7555
I0804 20:34:08.987832 140200711067520 basic_session_run_hooks.py:260] loss = 1.1059899, step = 118700 (3.149 sec)
I0804 20:34:12.123293 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8795
I0804 20:34:12.124708 140200711067520 basic_session_run_hooks.py:260] loss = 0.98225415, step = 118800 (3.137 sec)
I0804 20:34:15.277317 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7055
I0804 20:34:15.278626 140200711067520 basic_session_run_hooks.py:260] loss = 1.0587257, step = 118900 (3.154 sec)
I0804 20:34:18.470489 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 119000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:34:18.768795 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:34:18.809247 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3131
I0804 20:34:18.810371 140200711067520 basic_session_run_hooks.py:260] loss = 1.03945, step = 119000 (3.532 sec)
I0804 20:34:22.022519 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1213
I0804 20:34:22.023769 140200711067520 basic_session_run_hooks.py:260] loss = 1.0920397, step = 119100 (3.213 sec)
I0804 20:34:25.207462 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3977
I0804 20:34:25.208835 140200711067520 basic_session_run_hooks.py:260] loss = 1.0745568, step = 119200 (3.185 sec)
I0804 20:34:28.400153 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3215
I0804 20:34:28.401575 140200711067520 basic_session_run_hooks.py:260] loss = 1.1309339, step = 119300 (3.193 sec)
I0804 20:34:31.589653 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3527
I0804 20:34:31.590803 140200711067520 basic_session_run_hooks.py:260] loss = 1.1748291, step = 119400 (3.189 sec)
I0804 20:34:34.785667 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2891
I0804 20:34:34.786929 140200711067520 basic_session_run_hooks.py:260] loss = 1.0229839, step = 119500 (3.196 sec)
I0804 20:34:37.966528 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4382
I0804 20:34:37.967781 140200711067520 basic_session_run_hooks.py:260] loss = 1.0966163, step = 119600 (3.181 sec)
I0804 20:34:41.149518 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4171
I0804 20:34:41.150918 140200711067520 basic_session_run_hooks.py:260] loss = 1.0216304, step = 119700 (3.183 sec)
I0804 20:34:44.370516 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0463
I0804 20:34:44.371928 140200711067520 basic_session_run_hooks.py:260] loss = 1.0516363, step = 119800 (3.221 sec)
I0804 20:34:47.548163 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4695
I0804 20:34:47.549465 140200711067520 basic_session_run_hooks.py:260] loss = 1.1364056, step = 119900 (3.178 sec)
I0804 20:34:50.691684 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 120000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:34:50.983046 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:34:51.019909 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8037
I0804 20:34:51.020932 140200711067520 basic_session_run_hooks.py:260] loss = 1.0845891, step = 120000 (3.471 sec)
I0804 20:34:54.235029 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1034
I0804 20:34:54.236117 140200711067520 basic_session_run_hooks.py:260] loss = 1.0730962, step = 120100 (3.215 sec)
I0804 20:34:57.436091 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2395
I0804 20:34:57.437163 140200711067520 basic_session_run_hooks.py:260] loss = 1.0342497, step = 120200 (3.201 sec)
I0804 20:35:00.624269 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3658
I0804 20:35:00.625855 140200711067520 basic_session_run_hooks.py:260] loss = 1.1120515, step = 120300 (3.189 sec)
I0804 20:35:03.824599 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2471
I0804 20:35:03.825801 140200711067520 basic_session_run_hooks.py:260] loss = 1.2130164, step = 120400 (3.200 sec)
I0804 20:35:07.029782 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1992
I0804 20:35:07.031012 140200711067520 basic_session_run_hooks.py:260] loss = 1.064587, step = 120500 (3.205 sec)
I0804 20:35:10.299113 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5874
I0804 20:35:10.300337 140200711067520 basic_session_run_hooks.py:260] loss = 1.0807883, step = 120600 (3.269 sec)
I0804 20:35:13.537290 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8816
I0804 20:35:13.538964 140200711067520 basic_session_run_hooks.py:260] loss = 1.1965238, step = 120700 (3.239 sec)
I0804 20:35:16.756612 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0625
I0804 20:35:16.758248 140200711067520 basic_session_run_hooks.py:260] loss = 1.0814584, step = 120800 (3.219 sec)
I0804 20:35:19.966684 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1518
I0804 20:35:19.968061 140200711067520 basic_session_run_hooks.py:260] loss = 1.0949523, step = 120900 (3.210 sec)
I0804 20:35:23.117925 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 121000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:35:23.694339 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:35:23.738806 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 26.5101
I0804 20:35:23.739751 140200711067520 basic_session_run_hooks.py:260] loss = 1.0865313, step = 121000 (3.772 sec)
I0804 20:35:26.912853 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5059
I0804 20:35:26.914143 140200711067520 basic_session_run_hooks.py:260] loss = 1.0520372, step = 121100 (3.174 sec)
I0804 20:35:30.096608 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4094
I0804 20:35:30.098032 140200711067520 basic_session_run_hooks.py:260] loss = 1.1598045, step = 121200 (3.184 sec)
I0804 20:35:33.298439 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2323
I0804 20:35:33.299940 140200711067520 basic_session_run_hooks.py:260] loss = 1.0440681, step = 121300 (3.202 sec)
I0804 20:35:36.463465 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5954
I0804 20:35:36.464779 140200711067520 basic_session_run_hooks.py:260] loss = 1.0543853, step = 121400 (3.165 sec)
I0804 20:35:39.605806 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8231
I0804 20:35:39.607114 140200711067520 basic_session_run_hooks.py:260] loss = 1.1537057, step = 121500 (3.142 sec)
I0804 20:35:42.776291 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5409
I0804 20:35:42.777597 140200711067520 basic_session_run_hooks.py:260] loss = 1.0200535, step = 121600 (3.170 sec)
I0804 20:35:45.936746 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6412
I0804 20:35:45.938080 140200711067520 basic_session_run_hooks.py:260] loss = 1.0911199, step = 121700 (3.160 sec)
I0804 20:35:49.091807 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6949
I0804 20:35:49.093062 140200711067520 basic_session_run_hooks.py:260] loss = 1.0995606, step = 121800 (3.155 sec)
I0804 20:35:52.239125 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7732
I0804 20:35:52.240486 140200711067520 basic_session_run_hooks.py:260] loss = 1.0290549, step = 121900 (3.147 sec)
I0804 20:35:55.361014 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 122000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:35:55.653078 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:35:55.690761 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.9714
I0804 20:35:55.691773 140200711067520 basic_session_run_hooks.py:260] loss = 1.1479309, step = 122000 (3.451 sec)
I0804 20:35:58.859400 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5596
I0804 20:35:58.860771 140200711067520 basic_session_run_hooks.py:260] loss = 1.0614645, step = 122100 (3.169 sec)
I0804 20:36:02.046501 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3766
I0804 20:36:02.047888 140200711067520 basic_session_run_hooks.py:260] loss = 1.027438, step = 122200 (3.187 sec)
I0804 20:36:05.206586 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6447
I0804 20:36:05.207708 140200711067520 basic_session_run_hooks.py:260] loss = 1.1631184, step = 122300 (3.160 sec)
I0804 20:36:08.382392 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4879
I0804 20:36:08.383802 140200711067520 basic_session_run_hooks.py:260] loss = 1.0689884, step = 122400 (3.176 sec)
I0804 20:36:11.556793 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.502
I0804 20:36:11.558228 140200711067520 basic_session_run_hooks.py:260] loss = 1.0663255, step = 122500 (3.174 sec)
I0804 20:36:14.750292 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3136
I0804 20:36:14.751862 140200711067520 basic_session_run_hooks.py:260] loss = 1.0283654, step = 122600 (3.194 sec)
I0804 20:36:17.950440 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2487
I0804 20:36:17.951900 140200711067520 basic_session_run_hooks.py:260] loss = 1.0420986, step = 122700 (3.200 sec)
I0804 20:36:21.133486 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4164
I0804 20:36:21.134959 140200711067520 basic_session_run_hooks.py:260] loss = 0.99865633, step = 122800 (3.183 sec)
I0804 20:36:24.333903 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2459
I0804 20:36:24.335396 140200711067520 basic_session_run_hooks.py:260] loss = 1.0555323, step = 122900 (3.200 sec)
I0804 20:36:27.470313 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 123000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:36:27.780238 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 20:36:27.782696 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 20:36:27.927932 140200711067520 estimator.py:1145] Calling model_fn.
I0804 20:36:27.928857 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 20:36:27.929252 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 20:36:27.929347 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 20:36:27.929439 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 20:36:27.929508 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 20:36:27.929598 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 20:36:27.929668 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 20:36:28.015494 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 20:36:28.071400 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 20:36:28.212650 140200711067520 t2t_model.py:2172] Building model body
I0804 20:36:28.894059 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 20:36:29.791519 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 20:36:29.809734 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T20:36:29Z
I0804 20:36:29.977957 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 20:36:29.978564: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:36:29.978942: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 20:36:29.979021: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 20:36:29.979045: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 20:36:29.979070: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 20:36:29.979089: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 20:36:29.979108: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 20:36:29.979126: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 20:36:29.979146: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 20:36:29.979246: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:36:29.979698: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:36:29.980013: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 20:36:29.980054: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 20:36:29.980067: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 20:36:29.980080: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 20:36:29.980349: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:36:29.980781: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:36:29.981116: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 20:36:29.982384 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-123000
I0804 20:36:30.182147 140200711067520 session_manager.py:500] Running local_init_op.
I0804 20:36:30.225475 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 20:36:36.241044 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 20:36:41.555778 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 20:36:46.895450 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 20:36:52.205839 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 20:36:57.539475 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 20:37:02.878802 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 20:37:08.223222 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 20:37:13.581025 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 20:37:18.893887 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 20:37:23.718519 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-20:37:23
I0804 20:37:23.718739 140200711067520 estimator.py:2039] Saving dict for global step 123000: global_step = 123000, loss = 1.1774404, metrics-paper_generation_problem/targets/accuracy = 0.6735806, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8844384, metrics-paper_generation_problem/targets/approx_bleu_score = 0.48874897, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1774763, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.58298236, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.6953052
I0804 20:37:23.719188 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 123000: experiment/transformer/transformer_small/output/model.ckpt-123000
I0804 20:37:23.774665 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.68235
I0804 20:37:23.775649 140200711067520 basic_session_run_hooks.py:260] loss = 1.082283, step = 123000 (59.440 sec)
I0804 20:37:26.980385 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1944
I0804 20:37:26.981832 140200711067520 basic_session_run_hooks.py:260] loss = 1.0963401, step = 123100 (3.206 sec)
I0804 20:37:30.128124 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7688
I0804 20:37:30.129670 140200711067520 basic_session_run_hooks.py:260] loss = 1.1865437, step = 123200 (3.148 sec)
I0804 20:37:33.318118 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.348
I0804 20:37:33.319402 140200711067520 basic_session_run_hooks.py:260] loss = 1.0994161, step = 123300 (3.190 sec)
I0804 20:37:36.517031 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2606
I0804 20:37:36.518614 140200711067520 basic_session_run_hooks.py:260] loss = 1.1202749, step = 123400 (3.199 sec)
I0804 20:37:39.731846 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1061
I0804 20:37:39.733386 140200711067520 basic_session_run_hooks.py:260] loss = 1.1123505, step = 123500 (3.215 sec)
I0804 20:37:42.957886 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9976
I0804 20:37:42.959217 140200711067520 basic_session_run_hooks.py:260] loss = 1.1475307, step = 123600 (3.226 sec)
I0804 20:37:46.202774 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8178
I0804 20:37:46.204233 140200711067520 basic_session_run_hooks.py:260] loss = 1.03451, step = 123700 (3.245 sec)
I0804 20:37:49.452131 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7753
I0804 20:37:49.453565 140200711067520 basic_session_run_hooks.py:260] loss = 1.0477914, step = 123800 (3.249 sec)
I0804 20:37:52.674141 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0366
I0804 20:37:52.675581 140200711067520 basic_session_run_hooks.py:260] loss = 1.1100767, step = 123900 (3.222 sec)
I0804 20:37:55.838666 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 124000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:37:56.125630 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:37:56.172082 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5881
I0804 20:37:56.173300 140200711067520 basic_session_run_hooks.py:260] loss = 1.0994513, step = 124000 (3.498 sec)
I0804 20:37:59.391382 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0629
I0804 20:37:59.392782 140200711067520 basic_session_run_hooks.py:260] loss = 1.113581, step = 124100 (3.219 sec)
I0804 20:38:02.589250 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2705
I0804 20:38:02.590565 140200711067520 basic_session_run_hooks.py:260] loss = 1.1287934, step = 124200 (3.198 sec)
I0804 20:38:05.771501 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4245
I0804 20:38:05.772714 140200711067520 basic_session_run_hooks.py:260] loss = 1.0793189, step = 124300 (3.182 sec)
I0804 20:38:08.970206 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2626
I0804 20:38:08.971926 140200711067520 basic_session_run_hooks.py:260] loss = 1.113352, step = 124400 (3.199 sec)
I0804 20:38:12.194853 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0112
I0804 20:38:12.196446 140200711067520 basic_session_run_hooks.py:260] loss = 1.0449582, step = 124500 (3.224 sec)
I0804 20:38:15.382760 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3684
I0804 20:38:15.384160 140200711067520 basic_session_run_hooks.py:260] loss = 1.0987499, step = 124600 (3.188 sec)
I0804 20:38:18.561952 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4546
I0804 20:38:18.563073 140200711067520 basic_session_run_hooks.py:260] loss = 1.0843606, step = 124700 (3.179 sec)
I0804 20:38:21.743333 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4329
I0804 20:38:21.744897 140200711067520 basic_session_run_hooks.py:260] loss = 1.1422397, step = 124800 (3.182 sec)
I0804 20:38:24.938734 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2949
I0804 20:38:24.940100 140200711067520 basic_session_run_hooks.py:260] loss = 1.0729998, step = 124900 (3.195 sec)
I0804 20:38:28.083539 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 125000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:38:28.368057 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:38:28.414729 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7687
I0804 20:38:28.415873 140200711067520 basic_session_run_hooks.py:260] loss = 1.0956264, step = 125000 (3.476 sec)
I0804 20:38:31.620283 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.196
I0804 20:38:31.621807 140200711067520 basic_session_run_hooks.py:260] loss = 1.1172435, step = 125100 (3.206 sec)
I0804 20:38:34.778584 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6627
I0804 20:38:34.779774 140200711067520 basic_session_run_hooks.py:260] loss = 1.0724014, step = 125200 (3.158 sec)
I0804 20:38:37.974444 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2905
I0804 20:38:37.975730 140200711067520 basic_session_run_hooks.py:260] loss = 1.1266772, step = 125300 (3.196 sec)
I0804 20:38:41.142011 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5698
I0804 20:38:41.143534 140200711067520 basic_session_run_hooks.py:260] loss = 1.0674475, step = 125400 (3.168 sec)
I0804 20:38:44.319090 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4754
I0804 20:38:44.320341 140200711067520 basic_session_run_hooks.py:260] loss = 1.0427136, step = 125500 (3.177 sec)
I0804 20:38:47.496396 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4732
I0804 20:38:47.497910 140200711067520 basic_session_run_hooks.py:260] loss = 1.041305, step = 125600 (3.178 sec)
I0804 20:38:50.691324 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2996
I0804 20:38:50.692728 140200711067520 basic_session_run_hooks.py:260] loss = 1.0583968, step = 125700 (3.195 sec)
I0804 20:38:53.900554 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1603
I0804 20:38:53.901737 140200711067520 basic_session_run_hooks.py:260] loss = 1.1097583, step = 125800 (3.209 sec)
I0804 20:38:57.114171 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1174
I0804 20:38:57.115471 140200711067520 basic_session_run_hooks.py:260] loss = 1.0860314, step = 125900 (3.214 sec)
I0804 20:39:00.293199 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 126000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:39:00.601223 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:39:00.638583 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3734
I0804 20:39:00.639693 140200711067520 basic_session_run_hooks.py:260] loss = 1.0113444, step = 126000 (3.524 sec)
I0804 20:39:03.863777 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0061
I0804 20:39:03.865017 140200711067520 basic_session_run_hooks.py:260] loss = 1.0576187, step = 126100 (3.225 sec)
I0804 20:39:07.094885 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9491
I0804 20:39:07.096152 140200711067520 basic_session_run_hooks.py:260] loss = 1.0419586, step = 126200 (3.231 sec)
I0804 20:39:10.304655 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1549
I0804 20:39:10.305947 140200711067520 basic_session_run_hooks.py:260] loss = 1.0641183, step = 126300 (3.210 sec)
I0804 20:39:13.534946 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9569
I0804 20:39:13.536032 140200711067520 basic_session_run_hooks.py:260] loss = 1.0708435, step = 126400 (3.230 sec)
I0804 20:39:16.754787 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0575
I0804 20:39:16.755834 140200711067520 basic_session_run_hooks.py:260] loss = 1.0797784, step = 126500 (3.220 sec)
I0804 20:39:19.985444 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9539
I0804 20:39:19.986855 140200711067520 basic_session_run_hooks.py:260] loss = 1.1538894, step = 126600 (3.231 sec)
I0804 20:39:23.233272 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7895
I0804 20:39:23.234666 140200711067520 basic_session_run_hooks.py:260] loss = 1.072482, step = 126700 (3.248 sec)
I0804 20:39:26.445283 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1332
I0804 20:39:26.446735 140200711067520 basic_session_run_hooks.py:260] loss = 1.0325243, step = 126800 (3.212 sec)
I0804 20:39:29.604850 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6499
I0804 20:39:29.606286 140200711067520 basic_session_run_hooks.py:260] loss = 1.0868163, step = 126900 (3.160 sec)
I0804 20:39:32.754170 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 127000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:39:33.050257 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:39:33.094654 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6546
I0804 20:39:33.095911 140200711067520 basic_session_run_hooks.py:260] loss = 1.1500233, step = 127000 (3.490 sec)
I0804 20:39:36.245224 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7405
I0804 20:39:36.246477 140200711067520 basic_session_run_hooks.py:260] loss = 1.0908781, step = 127100 (3.151 sec)
I0804 20:39:39.421880 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4797
I0804 20:39:39.423111 140200711067520 basic_session_run_hooks.py:260] loss = 1.04847, step = 127200 (3.177 sec)
I0804 20:39:42.559155 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8749
I0804 20:39:42.560642 140200711067520 basic_session_run_hooks.py:260] loss = 1.0401886, step = 127300 (3.138 sec)
I0804 20:39:45.701390 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8244
I0804 20:39:45.702858 140200711067520 basic_session_run_hooks.py:260] loss = 1.0561467, step = 127400 (3.142 sec)
I0804 20:39:48.835510 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9072
I0804 20:39:48.836965 140200711067520 basic_session_run_hooks.py:260] loss = 1.1231767, step = 127500 (3.134 sec)
I0804 20:39:51.972291 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8795
I0804 20:39:51.974174 140200711067520 basic_session_run_hooks.py:260] loss = 1.0713776, step = 127600 (3.137 sec)
I0804 20:39:55.105963 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9114
I0804 20:39:55.107217 140200711067520 basic_session_run_hooks.py:260] loss = 1.0523237, step = 127700 (3.133 sec)
I0804 20:39:58.223631 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0753
I0804 20:39:58.225080 140200711067520 basic_session_run_hooks.py:260] loss = 1.0512531, step = 127800 (3.118 sec)
I0804 20:40:01.358133 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9031
I0804 20:40:01.359925 140200711067520 basic_session_run_hooks.py:260] loss = 1.0743198, step = 127900 (3.135 sec)
I0804 20:40:04.455983 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 128000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:40:04.755017 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:40:04.797125 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.078
I0804 20:40:04.798330 140200711067520 basic_session_run_hooks.py:260] loss = 1.0759904, step = 128000 (3.438 sec)
I0804 20:40:08.101009 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.2676
I0804 20:40:08.102454 140200711067520 basic_session_run_hooks.py:260] loss = 1.0607548, step = 128100 (3.304 sec)
I0804 20:40:11.276855 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4876
I0804 20:40:11.278177 140200711067520 basic_session_run_hooks.py:260] loss = 1.0402179, step = 128200 (3.176 sec)
I0804 20:40:14.484819 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1724
I0804 20:40:14.486127 140200711067520 basic_session_run_hooks.py:260] loss = 1.0919224, step = 128300 (3.208 sec)
I0804 20:40:17.669872 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3966
I0804 20:40:17.670987 140200711067520 basic_session_run_hooks.py:260] loss = 1.1169705, step = 128400 (3.185 sec)
I0804 20:40:20.887288 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0809
I0804 20:40:20.888721 140200711067520 basic_session_run_hooks.py:260] loss = 1.0930651, step = 128500 (3.218 sec)
I0804 20:40:24.144217 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7037
I0804 20:40:24.145595 140200711067520 basic_session_run_hooks.py:260] loss = 1.0856612, step = 128600 (3.257 sec)
I0804 20:40:27.370923 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9917
I0804 20:40:27.372296 140200711067520 basic_session_run_hooks.py:260] loss = 1.072879, step = 128700 (3.226 sec)
I0804 20:40:30.560835 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3486
I0804 20:40:30.562161 140200711067520 basic_session_run_hooks.py:260] loss = 1.0608481, step = 128800 (3.190 sec)
I0804 20:40:33.778152 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0817
I0804 20:40:33.779579 140200711067520 basic_session_run_hooks.py:260] loss = 0.970097, step = 128900 (3.217 sec)
I0804 20:40:36.947475 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 129000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:40:37.233327 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:40:37.267318 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.66
I0804 20:40:37.268455 140200711067520 basic_session_run_hooks.py:260] loss = 1.1035985, step = 129000 (3.489 sec)
I0804 20:40:40.476123 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1646
I0804 20:40:40.477338 140200711067520 basic_session_run_hooks.py:260] loss = 1.093757, step = 129100 (3.209 sec)
I0804 20:40:43.662528 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3835
I0804 20:40:43.663992 140200711067520 basic_session_run_hooks.py:260] loss = 1.0900975, step = 129200 (3.187 sec)
I0804 20:40:46.842132 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4502
I0804 20:40:46.843584 140200711067520 basic_session_run_hooks.py:260] loss = 0.9712583, step = 129300 (3.180 sec)
I0804 20:40:50.021458 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4541
I0804 20:40:50.022908 140200711067520 basic_session_run_hooks.py:260] loss = 1.0158708, step = 129400 (3.179 sec)
I0804 20:40:53.208156 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3796
I0804 20:40:53.209568 140200711067520 basic_session_run_hooks.py:260] loss = 1.0373573, step = 129500 (3.187 sec)
I0804 20:40:56.390100 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4273
I0804 20:40:56.391515 140200711067520 basic_session_run_hooks.py:260] loss = 1.0380429, step = 129600 (3.182 sec)
I0804 20:40:59.626668 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8968
I0804 20:40:59.627791 140200711067520 basic_session_run_hooks.py:260] loss = 1.0874157, step = 129700 (3.236 sec)
I0804 20:41:02.842219 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.099
I0804 20:41:02.843626 140200711067520 basic_session_run_hooks.py:260] loss = 1.1340221, step = 129800 (3.216 sec)
I0804 20:41:06.078336 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9011
I0804 20:41:06.080118 140200711067520 basic_session_run_hooks.py:260] loss = 1.1292241, step = 129900 (3.236 sec)
I0804 20:41:09.292046 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 130000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:41:09.580562 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:41:09.622297 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2168
I0804 20:41:09.623476 140200711067520 basic_session_run_hooks.py:260] loss = 1.0427296, step = 130000 (3.543 sec)
I0804 20:41:12.836399 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1131
I0804 20:41:12.837891 140200711067520 basic_session_run_hooks.py:260] loss = 1.1258874, step = 130100 (3.214 sec)
I0804 20:41:16.058569 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0354
I0804 20:41:16.059951 140200711067520 basic_session_run_hooks.py:260] loss = 1.1174846, step = 130200 (3.222 sec)
I0804 20:41:19.260151 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2344
I0804 20:41:19.261635 140200711067520 basic_session_run_hooks.py:260] loss = 1.0938197, step = 130300 (3.202 sec)
I0804 20:41:22.496538 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8988
I0804 20:41:22.497711 140200711067520 basic_session_run_hooks.py:260] loss = 1.1021628, step = 130400 (3.236 sec)
I0804 20:41:25.662663 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5841
I0804 20:41:25.664244 140200711067520 basic_session_run_hooks.py:260] loss = 1.0484191, step = 130500 (3.167 sec)
I0804 20:41:28.825593 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6165
I0804 20:41:28.826902 140200711067520 basic_session_run_hooks.py:260] loss = 1.0700041, step = 130600 (3.163 sec)
I0804 20:41:31.988309 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6183
I0804 20:41:31.989756 140200711067520 basic_session_run_hooks.py:260] loss = 1.0472919, step = 130700 (3.163 sec)
I0804 20:41:35.153801 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5905
I0804 20:41:35.155247 140200711067520 basic_session_run_hooks.py:260] loss = 1.0119936, step = 130800 (3.165 sec)
I0804 20:41:38.324057 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5435
I0804 20:41:38.325635 140200711067520 basic_session_run_hooks.py:260] loss = 1.0453254, step = 130900 (3.170 sec)
I0804 20:41:41.467307 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 131000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:41:41.762393 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:41:41.800677 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7632
I0804 20:41:41.801806 140200711067520 basic_session_run_hooks.py:260] loss = 1.1176745, step = 131000 (3.476 sec)
I0804 20:41:44.970807 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5448
I0804 20:41:44.973024 140200711067520 basic_session_run_hooks.py:260] loss = 1.0820979, step = 131100 (3.171 sec)
I0804 20:41:48.177493 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1849
I0804 20:41:48.178854 140200711067520 basic_session_run_hooks.py:260] loss = 1.1377584, step = 131200 (3.206 sec)
I0804 20:41:51.292839 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.099
I0804 20:41:51.293981 140200711067520 basic_session_run_hooks.py:260] loss = 1.1281378, step = 131300 (3.115 sec)
I0804 20:41:54.416601 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0129
I0804 20:41:54.418349 140200711067520 basic_session_run_hooks.py:260] loss = 1.0620182, step = 131400 (3.124 sec)
I0804 20:41:57.561689 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7956
I0804 20:41:57.563602 140200711067520 basic_session_run_hooks.py:260] loss = 1.0315901, step = 131500 (3.145 sec)
I0804 20:42:00.679996 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0685
I0804 20:42:00.681582 140200711067520 basic_session_run_hooks.py:260] loss = 1.0968261, step = 131600 (3.118 sec)
I0804 20:42:03.785217 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.2039
I0804 20:42:03.786567 140200711067520 basic_session_run_hooks.py:260] loss = 1.0644271, step = 131700 (3.105 sec)
I0804 20:42:06.909332 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0088
I0804 20:42:06.910666 140200711067520 basic_session_run_hooks.py:260] loss = 1.0871804, step = 131800 (3.124 sec)
I0804 20:42:10.047024 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8708
I0804 20:42:10.048594 140200711067520 basic_session_run_hooks.py:260] loss = 0.9765548, step = 131900 (3.138 sec)
I0804 20:42:13.158668 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 132000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:42:13.450074 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:42:13.491861 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0287
I0804 20:42:13.493029 140200711067520 basic_session_run_hooks.py:260] loss = 1.0469117, step = 132000 (3.444 sec)
I0804 20:42:16.680910 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3576
I0804 20:42:16.682149 140200711067520 basic_session_run_hooks.py:260] loss = 1.0413425, step = 132100 (3.189 sec)
I0804 20:42:19.854138 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5136
I0804 20:42:19.855494 140200711067520 basic_session_run_hooks.py:260] loss = 1.0608777, step = 132200 (3.173 sec)
I0804 20:42:22.991532 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8739
I0804 20:42:22.992773 140200711067520 basic_session_run_hooks.py:260] loss = 1.059728, step = 132300 (3.137 sec)
I0804 20:42:26.160691 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5537
I0804 20:42:26.161860 140200711067520 basic_session_run_hooks.py:260] loss = 1.0033706, step = 132400 (3.169 sec)
I0804 20:42:29.327187 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5806
I0804 20:42:29.328265 140200711067520 basic_session_run_hooks.py:260] loss = 1.1637979, step = 132500 (3.166 sec)
I0804 20:42:32.490544 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6125
I0804 20:42:32.491974 140200711067520 basic_session_run_hooks.py:260] loss = 1.1308662, step = 132600 (3.164 sec)
I0804 20:42:35.647413 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6767
I0804 20:42:35.648874 140200711067520 basic_session_run_hooks.py:260] loss = 1.1178335, step = 132700 (3.157 sec)
I0804 20:42:38.837271 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3495
I0804 20:42:38.838508 140200711067520 basic_session_run_hooks.py:260] loss = 1.0126661, step = 132800 (3.190 sec)
I0804 20:42:41.973653 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8837
I0804 20:42:41.974834 140200711067520 basic_session_run_hooks.py:260] loss = 1.0849683, step = 132900 (3.136 sec)
I0804 20:42:45.079466 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 133000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:42:45.370134 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:42:45.418967 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0247
I0804 20:42:45.420025 140200711067520 basic_session_run_hooks.py:260] loss = 1.1046358, step = 133000 (3.445 sec)
I0804 20:42:48.586327 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5725
I0804 20:42:48.587818 140200711067520 basic_session_run_hooks.py:260] loss = 1.1279885, step = 133100 (3.168 sec)
I0804 20:42:51.735360 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7557
I0804 20:42:51.737014 140200711067520 basic_session_run_hooks.py:260] loss = 1.0905515, step = 133200 (3.149 sec)
I0804 20:42:54.905623 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.543
I0804 20:42:54.906754 140200711067520 basic_session_run_hooks.py:260] loss = 1.0252862, step = 133300 (3.170 sec)
I0804 20:42:58.056308 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7393
I0804 20:42:58.057385 140200711067520 basic_session_run_hooks.py:260] loss = 1.1387486, step = 133400 (3.151 sec)
I0804 20:43:01.250044 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3113
I0804 20:43:01.251451 140200711067520 basic_session_run_hooks.py:260] loss = 1.0211879, step = 133500 (3.194 sec)
I0804 20:43:04.426708 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4796
I0804 20:43:04.428188 140200711067520 basic_session_run_hooks.py:260] loss = 1.1299152, step = 133600 (3.177 sec)
I0804 20:43:07.559190 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9234
I0804 20:43:07.560542 140200711067520 basic_session_run_hooks.py:260] loss = 1.050018, step = 133700 (3.132 sec)
I0804 20:43:10.696455 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.875
I0804 20:43:10.697635 140200711067520 basic_session_run_hooks.py:260] loss = 1.0229143, step = 133800 (3.137 sec)
I0804 20:43:13.827525 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9381
I0804 20:43:13.828680 140200711067520 basic_session_run_hooks.py:260] loss = 1.0428976, step = 133900 (3.131 sec)
I0804 20:43:16.900823 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 134000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:43:17.195328 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:43:17.238339 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.3182
I0804 20:43:17.239511 140200711067520 basic_session_run_hooks.py:260] loss = 1.0528085, step = 134000 (3.411 sec)
I0804 20:43:20.367578 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9572
I0804 20:43:20.369077 140200711067520 basic_session_run_hooks.py:260] loss = 1.0913424, step = 134100 (3.130 sec)
I0804 20:43:23.507946 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.843
I0804 20:43:23.509210 140200711067520 basic_session_run_hooks.py:260] loss = 1.0785948, step = 134200 (3.140 sec)
I0804 20:43:26.613680 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1985
I0804 20:43:26.614742 140200711067520 basic_session_run_hooks.py:260] loss = 1.0610332, step = 134300 (3.106 sec)
I0804 20:43:29.795215 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4313
I0804 20:43:29.796331 140200711067520 basic_session_run_hooks.py:260] loss = 1.1495882, step = 134400 (3.182 sec)
I0804 20:43:33.014469 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0633
I0804 20:43:33.015755 140200711067520 basic_session_run_hooks.py:260] loss = 1.0885842, step = 134500 (3.219 sec)
I0804 20:43:36.206177 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3311
I0804 20:43:36.207555 140200711067520 basic_session_run_hooks.py:260] loss = 1.0348818, step = 134600 (3.192 sec)
I0804 20:43:39.384307 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4651
I0804 20:43:39.386336 140200711067520 basic_session_run_hooks.py:260] loss = 1.0802544, step = 134700 (3.179 sec)
I0804 20:43:42.590806 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1866
I0804 20:43:42.592001 140200711067520 basic_session_run_hooks.py:260] loss = 1.0266082, step = 134800 (3.206 sec)
I0804 20:43:45.797454 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1854
I0804 20:43:45.798994 140200711067520 basic_session_run_hooks.py:260] loss = 1.1109791, step = 134900 (3.207 sec)
I0804 20:43:48.971472 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 135000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:43:49.259622 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:43:49.303150 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5246
I0804 20:43:49.304203 140200711067520 basic_session_run_hooks.py:260] loss = 1.0555471, step = 135000 (3.505 sec)
I0804 20:43:52.517617 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1096
I0804 20:43:52.518743 140200711067520 basic_session_run_hooks.py:260] loss = 1.0846152, step = 135100 (3.215 sec)
I0804 20:43:55.749151 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9452
I0804 20:43:55.750769 140200711067520 basic_session_run_hooks.py:260] loss = 1.0933, step = 135200 (3.232 sec)
I0804 20:43:58.953522 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2077
I0804 20:43:58.954869 140200711067520 basic_session_run_hooks.py:260] loss = 1.1077181, step = 135300 (3.204 sec)
I0804 20:44:02.159689 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1895
I0804 20:44:02.160859 140200711067520 basic_session_run_hooks.py:260] loss = 1.1138961, step = 135400 (3.206 sec)
I0804 20:44:05.379003 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0626
I0804 20:44:05.380405 140200711067520 basic_session_run_hooks.py:260] loss = 1.0261334, step = 135500 (3.220 sec)
I0804 20:44:08.585599 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1856
I0804 20:44:08.586899 140200711067520 basic_session_run_hooks.py:260] loss = 1.0855294, step = 135600 (3.206 sec)
I0804 20:44:11.792635 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1816
I0804 20:44:11.793835 140200711067520 basic_session_run_hooks.py:260] loss = 1.0330524, step = 135700 (3.207 sec)
I0804 20:44:15.021283 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9726
I0804 20:44:15.022676 140200711067520 basic_session_run_hooks.py:260] loss = 1.0865661, step = 135800 (3.229 sec)
I0804 20:44:18.227463 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1901
I0804 20:44:18.228676 140200711067520 basic_session_run_hooks.py:260] loss = 1.0364457, step = 135900 (3.206 sec)
I0804 20:44:21.387532 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 136000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:44:21.687354 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:44:21.726945 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5754
I0804 20:44:21.728139 140200711067520 basic_session_run_hooks.py:260] loss = 1.0794276, step = 136000 (3.499 sec)
I0804 20:44:24.870594 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8104
I0804 20:44:24.871739 140200711067520 basic_session_run_hooks.py:260] loss = 1.1802794, step = 136100 (3.144 sec)
I0804 20:44:28.012895 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8235
I0804 20:44:28.014011 140200711067520 basic_session_run_hooks.py:260] loss = 0.9795523, step = 136200 (3.142 sec)
I0804 20:44:31.140414 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9742
I0804 20:44:31.141723 140200711067520 basic_session_run_hooks.py:260] loss = 1.1664389, step = 136300 (3.128 sec)
I0804 20:44:34.267140 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9825
I0804 20:44:34.268611 140200711067520 basic_session_run_hooks.py:260] loss = 1.0758152, step = 136400 (3.127 sec)
I0804 20:44:37.400262 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9171
I0804 20:44:37.401856 140200711067520 basic_session_run_hooks.py:260] loss = 1.0803207, step = 136500 (3.133 sec)
I0804 20:44:40.524785 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0047
I0804 20:44:40.526043 140200711067520 basic_session_run_hooks.py:260] loss = 1.0325142, step = 136600 (3.124 sec)
I0804 20:44:43.679635 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6974
I0804 20:44:43.680903 140200711067520 basic_session_run_hooks.py:260] loss = 1.1149154, step = 136700 (3.155 sec)
I0804 20:44:46.864474 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3988
I0804 20:44:46.865813 140200711067520 basic_session_run_hooks.py:260] loss = 1.078677, step = 136800 (3.185 sec)
I0804 20:44:50.068979 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2061
I0804 20:44:50.070257 140200711067520 basic_session_run_hooks.py:260] loss = 0.9787005, step = 136900 (3.204 sec)
I0804 20:44:53.260771 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 137000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:44:53.563012 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:44:53.600328 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3175
I0804 20:44:53.601483 140200711067520 basic_session_run_hooks.py:260] loss = 1.0454419, step = 137000 (3.531 sec)
I0804 20:44:56.834106 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9238
I0804 20:44:56.835550 140200711067520 basic_session_run_hooks.py:260] loss = 1.0343883, step = 137100 (3.234 sec)
I0804 20:45:00.027345 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.316
I0804 20:45:00.029138 140200711067520 basic_session_run_hooks.py:260] loss = 1.160497, step = 137200 (3.194 sec)
I0804 20:45:03.242985 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0982
I0804 20:45:03.244071 140200711067520 basic_session_run_hooks.py:260] loss = 1.0587772, step = 137300 (3.215 sec)
I0804 20:45:06.445154 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2286
I0804 20:45:06.446624 140200711067520 basic_session_run_hooks.py:260] loss = 1.0848527, step = 137400 (3.203 sec)
I0804 20:45:09.623542 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4628
I0804 20:45:09.625025 140200711067520 basic_session_run_hooks.py:260] loss = 1.0918365, step = 137500 (3.178 sec)
I0804 20:45:12.774518 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7363
I0804 20:45:12.776092 140200711067520 basic_session_run_hooks.py:260] loss = 1.0217582, step = 137600 (3.151 sec)
I0804 20:45:15.907785 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9155
I0804 20:45:15.908856 140200711067520 basic_session_run_hooks.py:260] loss = 1.114336, step = 137700 (3.133 sec)
I0804 20:45:19.052369 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8005
I0804 20:45:19.053746 140200711067520 basic_session_run_hooks.py:260] loss = 1.1255345, step = 137800 (3.145 sec)
I0804 20:45:22.192118 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8497
I0804 20:45:22.193618 140200711067520 basic_session_run_hooks.py:260] loss = 1.1262385, step = 137900 (3.140 sec)
I0804 20:45:25.490690 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 138000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:45:25.785436 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:45:25.819191 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 27.5702
I0804 20:45:25.820817 140200711067520 basic_session_run_hooks.py:260] loss = 1.0668653, step = 138000 (3.627 sec)
I0804 20:45:29.055609 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8986
I0804 20:45:29.056716 140200711067520 basic_session_run_hooks.py:260] loss = 1.1204504, step = 138100 (3.236 sec)
I0804 20:45:32.298511 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8368
I0804 20:45:32.299721 140200711067520 basic_session_run_hooks.py:260] loss = 1.0449837, step = 138200 (3.243 sec)
I0804 20:45:35.529487 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9504
I0804 20:45:35.530717 140200711067520 basic_session_run_hooks.py:260] loss = 1.1321537, step = 138300 (3.231 sec)
I0804 20:45:38.779290 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7709
I0804 20:45:38.780524 140200711067520 basic_session_run_hooks.py:260] loss = 1.084304, step = 138400 (3.250 sec)
I0804 20:45:41.985563 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.189
I0804 20:45:41.987002 140200711067520 basic_session_run_hooks.py:260] loss = 1.1506865, step = 138500 (3.206 sec)
I0804 20:45:45.187072 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2352
I0804 20:45:45.188348 140200711067520 basic_session_run_hooks.py:260] loss = 1.1587224, step = 138600 (3.201 sec)
I0804 20:45:48.386053 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2599
I0804 20:45:48.387685 140200711067520 basic_session_run_hooks.py:260] loss = 1.0960499, step = 138700 (3.199 sec)
I0804 20:45:51.617053 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.95
I0804 20:45:51.618171 140200711067520 basic_session_run_hooks.py:260] loss = 1.0986222, step = 138800 (3.230 sec)
I0804 20:45:54.836978 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0568
I0804 20:45:54.838366 140200711067520 basic_session_run_hooks.py:260] loss = 1.05922, step = 138900 (3.220 sec)
I0804 20:45:58.031343 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 139000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:45:58.327249 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:45:58.364234 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3504
I0804 20:45:58.365210 140200711067520 basic_session_run_hooks.py:260] loss = 1.1003032, step = 139000 (3.527 sec)
I0804 20:46:01.568902 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2048
I0804 20:46:01.570155 140200711067520 basic_session_run_hooks.py:260] loss = 1.0675591, step = 139100 (3.205 sec)
I0804 20:46:04.764251 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2955
I0804 20:46:04.765499 140200711067520 basic_session_run_hooks.py:260] loss = 1.0923247, step = 139200 (3.195 sec)
I0804 20:46:07.981506 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0825
I0804 20:46:07.982943 140200711067520 basic_session_run_hooks.py:260] loss = 1.0267458, step = 139300 (3.217 sec)
I0804 20:46:11.190802 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1594
I0804 20:46:11.192162 140200711067520 basic_session_run_hooks.py:260] loss = 1.1827942, step = 139400 (3.209 sec)
I0804 20:46:14.412554 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0391
I0804 20:46:14.413913 140200711067520 basic_session_run_hooks.py:260] loss = 1.1147218, step = 139500 (3.222 sec)
I0804 20:46:17.650054 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8878
I0804 20:46:17.651351 140200711067520 basic_session_run_hooks.py:260] loss = 1.1053977, step = 139600 (3.237 sec)
I0804 20:46:20.877173 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9875
I0804 20:46:20.878671 140200711067520 basic_session_run_hooks.py:260] loss = 1.0931245, step = 139700 (3.227 sec)
I0804 20:46:24.069686 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3232
I0804 20:46:24.071011 140200711067520 basic_session_run_hooks.py:260] loss = 1.1093231, step = 139800 (3.192 sec)
I0804 20:46:27.264861 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2973
I0804 20:46:27.266047 140200711067520 basic_session_run_hooks.py:260] loss = 1.0340862, step = 139900 (3.195 sec)
I0804 20:46:30.431962 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 140000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:46:30.740548 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 20:46:30.741912 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 20:46:30.886035 140200711067520 estimator.py:1145] Calling model_fn.
I0804 20:46:30.887012 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 20:46:30.887408 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 20:46:30.887516 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 20:46:30.887597 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 20:46:30.887664 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 20:46:30.887745 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 20:46:30.887812 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 20:46:30.978196 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 20:46:31.041364 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 20:46:31.181602 140200711067520 t2t_model.py:2172] Building model body
I0804 20:46:32.132728 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 20:46:32.847655 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 20:46:32.865668 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T20:46:32Z
I0804 20:46:33.237009 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 20:46:33.237674: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:46:33.238078: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 20:46:33.238157: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 20:46:33.238181: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 20:46:33.238202: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 20:46:33.238226: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 20:46:33.238245: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 20:46:33.238264: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 20:46:33.238284: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 20:46:33.238378: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:46:33.238799: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:46:33.239127: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 20:46:33.239169: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 20:46:33.239182: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 20:46:33.239192: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 20:46:33.239479: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:46:33.239868: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:46:33.240199: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 20:46:33.241617 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-140000
I0804 20:46:33.442858 140200711067520 session_manager.py:500] Running local_init_op.
I0804 20:46:33.486590 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 20:46:39.456801 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 20:46:44.774641 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 20:46:50.067123 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 20:46:55.377333 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 20:47:00.722908 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 20:47:06.101334 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 20:47:11.409667 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 20:47:16.754611 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 20:47:22.056141 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 20:47:26.895255 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-20:47:26
I0804 20:47:26.895515 140200711067520 estimator.py:2039] Saving dict for global step 140000: global_step = 140000, loss = 1.1740515, metrics-paper_generation_problem/targets/accuracy = 0.6744676, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8845911, metrics-paper_generation_problem/targets/approx_bleu_score = 0.49061555, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1740876, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5846182, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.6961498
I0804 20:47:26.895967 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 140000: experiment/transformer/transformer_small/output/model.ckpt-140000
I0804 20:47:26.948387 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.6755
I0804 20:47:26.949515 140200711067520 basic_session_run_hooks.py:260] loss = 1.1040308, step = 140000 (59.683 sec)
I0804 20:47:30.183186 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9142
I0804 20:47:30.184460 140200711067520 basic_session_run_hooks.py:260] loss = 1.0124626, step = 140100 (3.235 sec)
I0804 20:47:33.392438 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1601
I0804 20:47:33.393904 140200711067520 basic_session_run_hooks.py:260] loss = 1.0635948, step = 140200 (3.209 sec)
I0804 20:47:36.641528 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7778
I0804 20:47:36.642913 140200711067520 basic_session_run_hooks.py:260] loss = 1.0940871, step = 140300 (3.249 sec)
I0804 20:47:39.865221 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0201
I0804 20:47:39.866847 140200711067520 basic_session_run_hooks.py:260] loss = 1.1815778, step = 140400 (3.224 sec)
I0804 20:47:43.084039 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0673
I0804 20:47:43.085795 140200711067520 basic_session_run_hooks.py:260] loss = 1.120026, step = 140500 (3.219 sec)
I0804 20:47:46.314089 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9593
I0804 20:47:46.315531 140200711067520 basic_session_run_hooks.py:260] loss = 1.0959108, step = 140600 (3.230 sec)
I0804 20:47:49.549001 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9129
I0804 20:47:49.550759 140200711067520 basic_session_run_hooks.py:260] loss = 1.0542556, step = 140700 (3.235 sec)
I0804 20:47:52.795512 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8023
I0804 20:47:52.796894 140200711067520 basic_session_run_hooks.py:260] loss = 1.116325, step = 140800 (3.246 sec)
I0804 20:47:56.025241 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9621
I0804 20:47:56.026617 140200711067520 basic_session_run_hooks.py:260] loss = 1.0791217, step = 140900 (3.230 sec)
I0804 20:47:59.207517 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 141000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:47:59.495932 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:47:59.537379 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4726
I0804 20:47:59.538526 140200711067520 basic_session_run_hooks.py:260] loss = 1.1619525, step = 141000 (3.512 sec)
I0804 20:48:02.777521 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8633
I0804 20:48:02.778805 140200711067520 basic_session_run_hooks.py:260] loss = 1.0559182, step = 141100 (3.240 sec)
I0804 20:48:05.945891 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5618
I0804 20:48:05.947291 140200711067520 basic_session_run_hooks.py:260] loss = 1.0496542, step = 141200 (3.168 sec)
I0804 20:48:09.106090 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6436
I0804 20:48:09.107568 140200711067520 basic_session_run_hooks.py:260] loss = 1.1584935, step = 141300 (3.160 sec)
I0804 20:48:12.303629 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2739
I0804 20:48:12.304892 140200711067520 basic_session_run_hooks.py:260] loss = 1.0148479, step = 141400 (3.197 sec)
I0804 20:48:15.497954 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3056
I0804 20:48:15.499299 140200711067520 basic_session_run_hooks.py:260] loss = 1.1368867, step = 141500 (3.194 sec)
I0804 20:48:18.707822 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.154
I0804 20:48:18.709301 140200711067520 basic_session_run_hooks.py:260] loss = 1.0708033, step = 141600 (3.210 sec)
I0804 20:48:21.887037 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4542
I0804 20:48:21.888110 140200711067520 basic_session_run_hooks.py:260] loss = 1.028151, step = 141700 (3.179 sec)
I0804 20:48:25.062543 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4913
I0804 20:48:25.064080 140200711067520 basic_session_run_hooks.py:260] loss = 1.0534006, step = 141800 (3.176 sec)
I0804 20:48:28.292749 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9576
I0804 20:48:28.294074 140200711067520 basic_session_run_hooks.py:260] loss = 1.1268613, step = 141900 (3.230 sec)
I0804 20:48:31.467974 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 142000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:48:31.762624 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:48:31.804103 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4789
I0804 20:48:31.805201 140200711067520 basic_session_run_hooks.py:260] loss = 1.0800323, step = 142000 (3.511 sec)
I0804 20:48:34.992795 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3609
I0804 20:48:34.993969 140200711067520 basic_session_run_hooks.py:260] loss = 1.0488505, step = 142100 (3.189 sec)
I0804 20:48:38.184164 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3345
I0804 20:48:38.185576 140200711067520 basic_session_run_hooks.py:260] loss = 1.077889, step = 142200 (3.192 sec)
I0804 20:48:41.385199 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.24
I0804 20:48:41.386526 140200711067520 basic_session_run_hooks.py:260] loss = 1.079634, step = 142300 (3.201 sec)
I0804 20:48:44.585571 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2465
I0804 20:48:44.586873 140200711067520 basic_session_run_hooks.py:260] loss = 1.0631179, step = 142400 (3.200 sec)
I0804 20:48:47.790556 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2014
I0804 20:48:47.791940 140200711067520 basic_session_run_hooks.py:260] loss = 1.1052933, step = 142500 (3.205 sec)
I0804 20:48:50.990255 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2528
I0804 20:48:50.991796 140200711067520 basic_session_run_hooks.py:260] loss = 1.1352735, step = 142600 (3.200 sec)
I0804 20:48:54.208864 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0692
I0804 20:48:54.210334 140200711067520 basic_session_run_hooks.py:260] loss = 1.0937493, step = 142700 (3.219 sec)
I0804 20:48:57.361183 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7228
I0804 20:48:57.362649 140200711067520 basic_session_run_hooks.py:260] loss = 1.0245173, step = 142800 (3.152 sec)
I0804 20:49:00.546635 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3927
I0804 20:49:00.548023 140200711067520 basic_session_run_hooks.py:260] loss = 1.0779155, step = 142900 (3.185 sec)
I0804 20:49:03.690277 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 143000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:49:03.981526 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:49:04.023592 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7605
I0804 20:49:04.024814 140200711067520 basic_session_run_hooks.py:260] loss = 1.0865078, step = 143000 (3.477 sec)
I0804 20:49:07.219273 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2925
I0804 20:49:07.220655 140200711067520 basic_session_run_hooks.py:260] loss = 1.0442033, step = 143100 (3.196 sec)
I0804 20:49:10.388193 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5564
I0804 20:49:10.389388 140200711067520 basic_session_run_hooks.py:260] loss = 1.1187522, step = 143200 (3.169 sec)
I0804 20:49:13.557383 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5539
I0804 20:49:13.558842 140200711067520 basic_session_run_hooks.py:260] loss = 1.0349634, step = 143300 (3.169 sec)
I0804 20:49:16.738564 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4351
I0804 20:49:16.740008 140200711067520 basic_session_run_hooks.py:260] loss = 1.0403154, step = 143400 (3.181 sec)
I0804 20:49:19.959482 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.047
I0804 20:49:19.961003 140200711067520 basic_session_run_hooks.py:260] loss = 1.0551487, step = 143500 (3.221 sec)
I0804 20:49:23.170873 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.139
I0804 20:49:23.172353 140200711067520 basic_session_run_hooks.py:260] loss = 0.99131703, step = 143600 (3.211 sec)
I0804 20:49:26.407843 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8931
I0804 20:49:26.409078 140200711067520 basic_session_run_hooks.py:260] loss = 1.0201955, step = 143700 (3.237 sec)
I0804 20:49:29.624135 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0916
I0804 20:49:29.625540 140200711067520 basic_session_run_hooks.py:260] loss = 0.9983866, step = 143800 (3.216 sec)
I0804 20:49:32.819470 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2959
I0804 20:49:32.821713 140200711067520 basic_session_run_hooks.py:260] loss = 1.0695317, step = 143900 (3.196 sec)
I0804 20:49:35.996216 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 144000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:49:36.287481 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:49:36.335020 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4446
I0804 20:49:36.336229 140200711067520 basic_session_run_hooks.py:260] loss = 1.0999237, step = 144000 (3.515 sec)
I0804 20:49:39.543190 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1707
I0804 20:49:39.544354 140200711067520 basic_session_run_hooks.py:260] loss = 1.0920385, step = 144100 (3.208 sec)
I0804 20:49:42.741966 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2619
I0804 20:49:42.743325 140200711067520 basic_session_run_hooks.py:260] loss = 1.1168659, step = 144200 (3.199 sec)
I0804 20:49:45.977533 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9068
I0804 20:49:45.978854 140200711067520 basic_session_run_hooks.py:260] loss = 1.0663586, step = 144300 (3.236 sec)
I0804 20:49:49.205698 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.977
I0804 20:49:49.207030 140200711067520 basic_session_run_hooks.py:260] loss = 1.1064819, step = 144400 (3.228 sec)
I0804 20:49:52.395147 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3535
I0804 20:49:52.396846 140200711067520 basic_session_run_hooks.py:260] loss = 1.1119072, step = 144500 (3.190 sec)
I0804 20:49:55.597608 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2259
I0804 20:49:55.599075 140200711067520 basic_session_run_hooks.py:260] loss = 1.0011477, step = 144600 (3.202 sec)
I0804 20:49:58.784530 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3788
I0804 20:49:58.786254 140200711067520 basic_session_run_hooks.py:260] loss = 1.0623722, step = 144700 (3.187 sec)
I0804 20:50:01.970008 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.392
I0804 20:50:01.971492 140200711067520 basic_session_run_hooks.py:260] loss = 1.0437233, step = 144800 (3.185 sec)
I0804 20:50:05.175004 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2013
I0804 20:50:05.176366 140200711067520 basic_session_run_hooks.py:260] loss = 1.0604427, step = 144900 (3.205 sec)
I0804 20:50:08.354514 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 145000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:50:08.642878 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:50:08.680702 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5249
I0804 20:50:08.683180 140200711067520 basic_session_run_hooks.py:260] loss = 1.0724804, step = 145000 (3.507 sec)
I0804 20:50:11.902224 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0413
I0804 20:50:11.903615 140200711067520 basic_session_run_hooks.py:260] loss = 1.1456715, step = 145100 (3.220 sec)
I0804 20:50:15.090924 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3607
I0804 20:50:15.092305 140200711067520 basic_session_run_hooks.py:260] loss = 1.0799633, step = 145200 (3.189 sec)
I0804 20:50:18.260846 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5465
I0804 20:50:18.262291 140200711067520 basic_session_run_hooks.py:260] loss = 1.0913296, step = 145300 (3.170 sec)
I0804 20:50:21.455857 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2989
I0804 20:50:21.457474 140200711067520 basic_session_run_hooks.py:260] loss = 1.0079027, step = 145400 (3.195 sec)
I0804 20:50:24.633621 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4687
I0804 20:50:24.634920 140200711067520 basic_session_run_hooks.py:260] loss = 1.0368984, step = 145500 (3.177 sec)
I0804 20:50:27.827409 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3113
I0804 20:50:27.828638 140200711067520 basic_session_run_hooks.py:260] loss = 1.0572461, step = 145600 (3.194 sec)
I0804 20:50:31.005746 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4623
I0804 20:50:31.006844 140200711067520 basic_session_run_hooks.py:260] loss = 1.0328832, step = 145700 (3.178 sec)
I0804 20:50:34.239774 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9212
I0804 20:50:34.241171 140200711067520 basic_session_run_hooks.py:260] loss = 1.0589328, step = 145800 (3.234 sec)
I0804 20:50:37.476244 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8979
I0804 20:50:37.477594 140200711067520 basic_session_run_hooks.py:260] loss = 1.0393786, step = 145900 (3.236 sec)
I0804 20:50:40.675087 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 146000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:50:40.980960 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:50:41.021521 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2065
I0804 20:50:41.022607 140200711067520 basic_session_run_hooks.py:260] loss = 1.1042863, step = 146000 (3.545 sec)
I0804 20:50:44.294685 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5516
I0804 20:50:44.295788 140200711067520 basic_session_run_hooks.py:260] loss = 1.0853151, step = 146100 (3.273 sec)
I0804 20:50:47.556083 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6617
I0804 20:50:47.557283 140200711067520 basic_session_run_hooks.py:260] loss = 1.1619797, step = 146200 (3.261 sec)
I0804 20:50:50.798356 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8426
I0804 20:50:50.799697 140200711067520 basic_session_run_hooks.py:260] loss = 1.1011312, step = 146300 (3.242 sec)
I0804 20:50:54.018729 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0526
I0804 20:50:54.019832 140200711067520 basic_session_run_hooks.py:260] loss = 0.9940027, step = 146400 (3.220 sec)
I0804 20:50:57.214328 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2926
I0804 20:50:57.215503 140200711067520 basic_session_run_hooks.py:260] loss = 1.066598, step = 146500 (3.196 sec)
I0804 20:51:00.443391 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9688
I0804 20:51:00.444790 140200711067520 basic_session_run_hooks.py:260] loss = 1.1141264, step = 146600 (3.229 sec)
I0804 20:51:03.653964 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1471
I0804 20:51:03.655304 140200711067520 basic_session_run_hooks.py:260] loss = 1.0237797, step = 146700 (3.211 sec)
I0804 20:51:06.853130 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2583
I0804 20:51:06.854560 140200711067520 basic_session_run_hooks.py:260] loss = 1.1441579, step = 146800 (3.199 sec)
I0804 20:51:10.046369 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.316
I0804 20:51:10.047856 140200711067520 basic_session_run_hooks.py:260] loss = 1.1008765, step = 146900 (3.193 sec)
I0804 20:51:13.225974 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 147000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:51:13.521635 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:51:13.561985 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4445
I0804 20:51:13.563014 140200711067520 basic_session_run_hooks.py:260] loss = 1.0718569, step = 147000 (3.515 sec)
I0804 20:51:16.719475 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.671
I0804 20:51:16.720761 140200711067520 basic_session_run_hooks.py:260] loss = 1.0063621, step = 147100 (3.158 sec)
I0804 20:51:19.923353 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2123
I0804 20:51:19.924782 140200711067520 basic_session_run_hooks.py:260] loss = 1.0325418, step = 147200 (3.204 sec)
I0804 20:51:23.105564 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4248
I0804 20:51:23.106847 140200711067520 basic_session_run_hooks.py:260] loss = 1.0429054, step = 147300 (3.182 sec)
I0804 20:51:26.305260 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2528
I0804 20:51:26.306668 140200711067520 basic_session_run_hooks.py:260] loss = 1.1683332, step = 147400 (3.200 sec)
I0804 20:51:29.507179 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2312
I0804 20:51:29.508507 140200711067520 basic_session_run_hooks.py:260] loss = 1.0877167, step = 147500 (3.202 sec)
I0804 20:51:32.716156 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1627
I0804 20:51:32.717607 140200711067520 basic_session_run_hooks.py:260] loss = 1.0538639, step = 147600 (3.209 sec)
I0804 20:51:35.908444 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3255
I0804 20:51:35.909848 140200711067520 basic_session_run_hooks.py:260] loss = 1.0727574, step = 147700 (3.192 sec)
I0804 20:51:39.069797 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6318
I0804 20:51:39.070884 140200711067520 basic_session_run_hooks.py:260] loss = 1.0426818, step = 147800 (3.161 sec)
I0804 20:51:42.366912 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.3296
I0804 20:51:42.368171 140200711067520 basic_session_run_hooks.py:260] loss = 0.99297297, step = 147900 (3.297 sec)
I0804 20:51:45.521033 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 148000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:51:45.814373 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:51:45.857338 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6496
I0804 20:51:45.858448 140200711067520 basic_session_run_hooks.py:260] loss = 1.0729353, step = 148000 (3.490 sec)
I0804 20:51:49.047729 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3446
I0804 20:51:49.049293 140200711067520 basic_session_run_hooks.py:260] loss = 1.0904833, step = 148100 (3.191 sec)
I0804 20:51:52.216555 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5576
I0804 20:51:52.217751 140200711067520 basic_session_run_hooks.py:260] loss = 0.9923338, step = 148200 (3.168 sec)
I0804 20:51:55.386010 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5509
I0804 20:51:55.387530 140200711067520 basic_session_run_hooks.py:260] loss = 1.0557016, step = 148300 (3.170 sec)
I0804 20:51:58.541461 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6913
I0804 20:51:58.542615 140200711067520 basic_session_run_hooks.py:260] loss = 1.0687994, step = 148400 (3.155 sec)
I0804 20:52:01.710777 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5524
I0804 20:52:01.712093 140200711067520 basic_session_run_hooks.py:260] loss = 1.1339226, step = 148500 (3.169 sec)
I0804 20:52:04.868438 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6691
I0804 20:52:04.869797 140200711067520 basic_session_run_hooks.py:260] loss = 1.0378462, step = 148600 (3.158 sec)
I0804 20:52:08.119814 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7561
I0804 20:52:08.120959 140200711067520 basic_session_run_hooks.py:260] loss = 1.0255045, step = 148700 (3.251 sec)
I0804 20:52:11.341343 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0413
I0804 20:52:11.342772 140200711067520 basic_session_run_hooks.py:260] loss = 1.02949, step = 148800 (3.222 sec)
I0804 20:52:14.584829 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.831
I0804 20:52:14.586122 140200711067520 basic_session_run_hooks.py:260] loss = 1.0511491, step = 148900 (3.243 sec)
I0804 20:52:17.770954 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 149000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:52:18.063652 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:52:18.107100 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3909
I0804 20:52:18.108300 140200711067520 basic_session_run_hooks.py:260] loss = 1.098434, step = 149000 (3.522 sec)
I0804 20:52:21.345333 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8807
I0804 20:52:21.346591 140200711067520 basic_session_run_hooks.py:260] loss = 1.0896103, step = 149100 (3.238 sec)
I0804 20:52:24.576198 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9516
I0804 20:52:24.577336 140200711067520 basic_session_run_hooks.py:260] loss = 1.0783967, step = 149200 (3.231 sec)
I0804 20:52:27.795012 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0674
I0804 20:52:27.796095 140200711067520 basic_session_run_hooks.py:260] loss = 1.1130404, step = 149300 (3.219 sec)
I0804 20:52:31.024897 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9608
I0804 20:52:31.026270 140200711067520 basic_session_run_hooks.py:260] loss = 1.0087849, step = 149400 (3.230 sec)
I0804 20:52:34.224781 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.251
I0804 20:52:34.226246 140200711067520 basic_session_run_hooks.py:260] loss = 1.0532873, step = 149500 (3.200 sec)
I0804 20:52:37.391507 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5786
I0804 20:52:37.392726 140200711067520 basic_session_run_hooks.py:260] loss = 1.0322157, step = 149600 (3.166 sec)
I0804 20:52:40.553969 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6208
I0804 20:52:40.554995 140200711067520 basic_session_run_hooks.py:260] loss = 1.0899488, step = 149700 (3.162 sec)
I0804 20:52:43.736306 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4236
I0804 20:52:43.737888 140200711067520 basic_session_run_hooks.py:260] loss = 1.171801, step = 149800 (3.183 sec)
I0804 20:52:46.916042 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.449
I0804 20:52:46.917962 140200711067520 basic_session_run_hooks.py:260] loss = 1.0214597, step = 149900 (3.180 sec)
I0804 20:52:50.061162 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 150000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:52:50.358166 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:52:50.398960 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7114
I0804 20:52:50.400115 140200711067520 basic_session_run_hooks.py:260] loss = 1.0841259, step = 150000 (3.482 sec)
I0804 20:52:53.598326 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2565
I0804 20:52:53.600679 140200711067520 basic_session_run_hooks.py:260] loss = 1.076192, step = 150100 (3.201 sec)
I0804 20:52:56.798204 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.251
I0804 20:52:56.799676 140200711067520 basic_session_run_hooks.py:260] loss = 1.0375971, step = 150200 (3.199 sec)
I0804 20:53:00.004552 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1885
I0804 20:53:00.005941 140200711067520 basic_session_run_hooks.py:260] loss = 1.1015152, step = 150300 (3.206 sec)
I0804 20:53:03.176701 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5243
I0804 20:53:03.177894 140200711067520 basic_session_run_hooks.py:260] loss = 1.0445292, step = 150400 (3.172 sec)
I0804 20:53:06.368019 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3349
I0804 20:53:06.369090 140200711067520 basic_session_run_hooks.py:260] loss = 1.0518298, step = 150500 (3.191 sec)
I0804 20:53:09.557311 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3549
I0804 20:53:09.558546 140200711067520 basic_session_run_hooks.py:260] loss = 1.1007837, step = 150600 (3.189 sec)
I0804 20:53:12.748851 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3328
I0804 20:53:12.750516 140200711067520 basic_session_run_hooks.py:260] loss = 1.0668285, step = 150700 (3.192 sec)
I0804 20:53:15.944797 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2898
I0804 20:53:15.946507 140200711067520 basic_session_run_hooks.py:260] loss = 1.0991299, step = 150800 (3.196 sec)
I0804 20:53:19.131592 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3794
I0804 20:53:19.132844 140200711067520 basic_session_run_hooks.py:260] loss = 1.1489006, step = 150900 (3.186 sec)
I0804 20:53:22.333837 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 151000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:53:22.625712 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:53:22.666099 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2922
I0804 20:53:22.667282 140200711067520 basic_session_run_hooks.py:260] loss = 1.1180941, step = 151000 (3.534 sec)
I0804 20:53:25.892325 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9962
I0804 20:53:25.893622 140200711067520 basic_session_run_hooks.py:260] loss = 1.1482745, step = 151100 (3.226 sec)
I0804 20:53:29.118964 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.992
I0804 20:53:29.120629 140200711067520 basic_session_run_hooks.py:260] loss = 1.0484141, step = 151200 (3.227 sec)
I0804 20:53:32.332165 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1216
I0804 20:53:32.333396 140200711067520 basic_session_run_hooks.py:260] loss = 1.0323137, step = 151300 (3.213 sec)
I0804 20:53:35.554987 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0287
I0804 20:53:35.556530 140200711067520 basic_session_run_hooks.py:260] loss = 1.070409, step = 151400 (3.223 sec)
I0804 20:53:38.780508 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0029
I0804 20:53:38.781985 140200711067520 basic_session_run_hooks.py:260] loss = 0.9916365, step = 151500 (3.225 sec)
I0804 20:53:42.027917 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7938
I0804 20:53:42.029247 140200711067520 basic_session_run_hooks.py:260] loss = 1.0625739, step = 151600 (3.247 sec)
I0804 20:53:45.265061 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8914
I0804 20:53:45.266512 140200711067520 basic_session_run_hooks.py:260] loss = 1.0853715, step = 151700 (3.237 sec)
I0804 20:53:48.508253 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8337
I0804 20:53:48.509379 140200711067520 basic_session_run_hooks.py:260] loss = 1.1071123, step = 151800 (3.243 sec)
I0804 20:53:51.677837 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.55
I0804 20:53:51.679147 140200711067520 basic_session_run_hooks.py:260] loss = 1.0069091, step = 151900 (3.170 sec)
I0804 20:53:54.841510 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 152000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:53:55.147358 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:53:55.185669 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5075
I0804 20:53:55.186586 140200711067520 basic_session_run_hooks.py:260] loss = 1.0763937, step = 152000 (3.507 sec)
I0804 20:53:58.341211 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6906
I0804 20:53:58.342822 140200711067520 basic_session_run_hooks.py:260] loss = 0.99667484, step = 152100 (3.156 sec)
I0804 20:54:01.500623 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6515
I0804 20:54:01.501791 140200711067520 basic_session_run_hooks.py:260] loss = 1.0662063, step = 152200 (3.159 sec)
I0804 20:54:04.643302 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.82
I0804 20:54:04.644603 140200711067520 basic_session_run_hooks.py:260] loss = 1.1200616, step = 152300 (3.143 sec)
I0804 20:54:07.799272 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6858
I0804 20:54:07.800891 140200711067520 basic_session_run_hooks.py:260] loss = 1.1827819, step = 152400 (3.156 sec)
I0804 20:54:10.960974 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6285
I0804 20:54:10.962157 140200711067520 basic_session_run_hooks.py:260] loss = 1.0336945, step = 152500 (3.161 sec)
I0804 20:54:14.201079 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8633
I0804 20:54:14.202256 140200711067520 basic_session_run_hooks.py:260] loss = 1.074345, step = 152600 (3.240 sec)
I0804 20:54:17.414321 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1212
I0804 20:54:17.415876 140200711067520 basic_session_run_hooks.py:260] loss = 1.1291722, step = 152700 (3.214 sec)
I0804 20:54:20.590504 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4846
I0804 20:54:20.591771 140200711067520 basic_session_run_hooks.py:260] loss = 1.0390936, step = 152800 (3.176 sec)
I0804 20:54:23.798376 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1731
I0804 20:54:23.799653 140200711067520 basic_session_run_hooks.py:260] loss = 1.1830604, step = 152900 (3.208 sec)
I0804 20:54:26.945442 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 153000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:54:27.233982 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:54:27.276478 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7513
I0804 20:54:27.277547 140200711067520 basic_session_run_hooks.py:260] loss = 1.0572169, step = 153000 (3.478 sec)
I0804 20:54:30.468331 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3299
I0804 20:54:30.469610 140200711067520 basic_session_run_hooks.py:260] loss = 1.0615433, step = 153100 (3.192 sec)
I0804 20:54:33.672699 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2074
I0804 20:54:33.673935 140200711067520 basic_session_run_hooks.py:260] loss = 1.103894, step = 153200 (3.204 sec)
I0804 20:54:36.866636 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3092
I0804 20:54:36.867805 140200711067520 basic_session_run_hooks.py:260] loss = 1.0348737, step = 153300 (3.194 sec)
I0804 20:54:40.086945 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.053
I0804 20:54:40.088098 140200711067520 basic_session_run_hooks.py:260] loss = 1.0602078, step = 153400 (3.220 sec)
I0804 20:54:43.315766 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.971
I0804 20:54:43.317113 140200711067520 basic_session_run_hooks.py:260] loss = 1.079171, step = 153500 (3.229 sec)
I0804 20:54:46.511033 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2964
I0804 20:54:46.512761 140200711067520 basic_session_run_hooks.py:260] loss = 1.1007608, step = 153600 (3.196 sec)
I0804 20:54:49.708387 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.276
I0804 20:54:49.709545 140200711067520 basic_session_run_hooks.py:260] loss = 1.1010245, step = 153700 (3.197 sec)
I0804 20:54:52.914182 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1934
I0804 20:54:52.915485 140200711067520 basic_session_run_hooks.py:260] loss = 1.1531675, step = 153800 (3.206 sec)
I0804 20:54:56.126117 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1339
I0804 20:54:56.127387 140200711067520 basic_session_run_hooks.py:260] loss = 1.0913765, step = 153900 (3.212 sec)
I0804 20:54:59.313906 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 154000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:54:59.612089 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:54:59.649467 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3841
I0804 20:54:59.650512 140200711067520 basic_session_run_hooks.py:260] loss = 1.0208151, step = 154000 (3.523 sec)
I0804 20:55:02.867242 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0749
I0804 20:55:02.868443 140200711067520 basic_session_run_hooks.py:260] loss = 1.115311, step = 154100 (3.218 sec)
I0804 20:55:06.115042 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7903
I0804 20:55:06.116167 140200711067520 basic_session_run_hooks.py:260] loss = 1.0488889, step = 154200 (3.248 sec)
I0804 20:55:09.307703 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3218
I0804 20:55:09.309012 140200711067520 basic_session_run_hooks.py:260] loss = 1.0031601, step = 154300 (3.193 sec)
I0804 20:55:12.486435 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.459
I0804 20:55:12.487813 140200711067520 basic_session_run_hooks.py:260] loss = 1.0530965, step = 154400 (3.179 sec)
I0804 20:55:15.661252 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4977
I0804 20:55:15.662985 140200711067520 basic_session_run_hooks.py:260] loss = 1.0515256, step = 154500 (3.175 sec)
I0804 20:55:18.838988 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4691
I0804 20:55:18.840595 140200711067520 basic_session_run_hooks.py:260] loss = 1.0675071, step = 154600 (3.178 sec)
I0804 20:55:22.000356 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6318
I0804 20:55:22.001825 140200711067520 basic_session_run_hooks.py:260] loss = 1.0666847, step = 154700 (3.161 sec)
I0804 20:55:25.187623 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.375
I0804 20:55:25.188810 140200711067520 basic_session_run_hooks.py:260] loss = 1.1048673, step = 154800 (3.187 sec)
I0804 20:55:28.347484 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.647
I0804 20:55:28.348901 140200711067520 basic_session_run_hooks.py:260] loss = 0.9542732, step = 154900 (3.160 sec)
I0804 20:55:31.529276 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 155000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:55:31.818170 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:55:31.857296 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4913
I0804 20:55:31.858466 140200711067520 basic_session_run_hooks.py:260] loss = 1.143069, step = 155000 (3.510 sec)
I0804 20:55:35.059585 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2281
I0804 20:55:35.060985 140200711067520 basic_session_run_hooks.py:260] loss = 1.024289, step = 155100 (3.203 sec)
I0804 20:55:38.264364 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2032
I0804 20:55:38.265714 140200711067520 basic_session_run_hooks.py:260] loss = 1.0517182, step = 155200 (3.205 sec)
I0804 20:55:41.471854 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1772
I0804 20:55:41.473152 140200711067520 basic_session_run_hooks.py:260] loss = 1.0343468, step = 155300 (3.207 sec)
I0804 20:55:44.682853 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1428
I0804 20:55:44.683967 140200711067520 basic_session_run_hooks.py:260] loss = 1.0630282, step = 155400 (3.211 sec)
I0804 20:55:47.885851 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2207
I0804 20:55:47.887028 140200711067520 basic_session_run_hooks.py:260] loss = 1.0826725, step = 155500 (3.203 sec)
I0804 20:55:51.096087 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1506
I0804 20:55:51.097315 140200711067520 basic_session_run_hooks.py:260] loss = 1.0631969, step = 155600 (3.210 sec)
I0804 20:55:54.365559 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5859
I0804 20:55:54.366886 140200711067520 basic_session_run_hooks.py:260] loss = 0.9837358, step = 155700 (3.270 sec)
I0804 20:55:57.601352 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9041
I0804 20:55:57.602503 140200711067520 basic_session_run_hooks.py:260] loss = 1.0632521, step = 155800 (3.236 sec)
I0804 20:56:00.823688 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0333
I0804 20:56:00.824825 140200711067520 basic_session_run_hooks.py:260] loss = 1.0438957, step = 155900 (3.222 sec)
I0804 20:56:03.968075 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 156000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:56:04.267440 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:56:04.308106 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6992
I0804 20:56:04.309144 140200711067520 basic_session_run_hooks.py:260] loss = 1.1873302, step = 156000 (3.484 sec)
I0804 20:56:07.517452 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1593
I0804 20:56:07.518944 140200711067520 basic_session_run_hooks.py:260] loss = 1.0715199, step = 156100 (3.210 sec)
I0804 20:56:10.721178 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2136
I0804 20:56:10.722690 140200711067520 basic_session_run_hooks.py:260] loss = 1.2598089, step = 156200 (3.204 sec)
I0804 20:56:13.935627 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1095
I0804 20:56:13.937036 140200711067520 basic_session_run_hooks.py:260] loss = 1.1144172, step = 156300 (3.214 sec)
I0804 20:56:17.173557 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8841
I0804 20:56:17.174811 140200711067520 basic_session_run_hooks.py:260] loss = 1.005496, step = 156400 (3.238 sec)
I0804 20:56:20.402896 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.966
I0804 20:56:20.404345 140200711067520 basic_session_run_hooks.py:260] loss = 1.0726296, step = 156500 (3.230 sec)
I0804 20:56:23.642033 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8724
I0804 20:56:23.643634 140200711067520 basic_session_run_hooks.py:260] loss = 1.1007835, step = 156600 (3.239 sec)
I0804 20:56:26.873818 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9425
I0804 20:56:26.875308 140200711067520 basic_session_run_hooks.py:260] loss = 1.0257366, step = 156700 (3.232 sec)
I0804 20:56:30.088623 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1061
I0804 20:56:30.090245 140200711067520 basic_session_run_hooks.py:260] loss = 1.0797359, step = 156800 (3.215 sec)
I0804 20:56:33.298161 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1572
I0804 20:56:33.299257 140200711067520 basic_session_run_hooks.py:260] loss = 1.0652769, step = 156900 (3.209 sec)
I0804 20:56:36.498858 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 157000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:56:36.796658 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 20:56:36.798084 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 20:56:36.949212 140200711067520 estimator.py:1145] Calling model_fn.
I0804 20:56:36.950166 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 20:56:36.950587 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 20:56:36.950681 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 20:56:36.950762 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 20:56:36.950836 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 20:56:36.950922 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 20:56:36.950989 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 20:56:37.040938 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 20:56:37.096696 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 20:56:37.243588 140200711067520 t2t_model.py:2172] Building model body
I0804 20:56:38.202975 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 20:56:38.905867 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 20:56:38.923975 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T20:56:38Z
I0804 20:56:39.098509 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 20:56:39.099106: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:56:39.099518: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 20:56:39.099665: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 20:56:39.099693: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 20:56:39.099714: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 20:56:39.099734: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 20:56:39.099753: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 20:56:39.099771: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 20:56:39.099792: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 20:56:39.099900: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:56:39.100296: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:56:39.100635: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 20:56:39.100677: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 20:56:39.100691: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 20:56:39.100702: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 20:56:39.100991: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:56:39.101388: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 20:56:39.101728: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 20:56:39.103050 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-157000
I0804 20:56:39.312135 140200711067520 session_manager.py:500] Running local_init_op.
I0804 20:56:39.359546 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 20:56:45.365514 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 20:56:50.668572 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 20:56:55.991714 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 20:57:01.333204 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 20:57:06.656862 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 20:57:12.001010 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 20:57:17.299221 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 20:57:22.616360 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 20:57:27.940987 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 20:57:32.757968 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-20:57:32
I0804 20:57:32.758194 140200711067520 estimator.py:2039] Saving dict for global step 157000: global_step = 157000, loss = 1.1703839, metrics-paper_generation_problem/targets/accuracy = 0.67514414, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.88518596, metrics-paper_generation_problem/targets/approx_bleu_score = 0.4908059, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1704199, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5853118, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.6966693
I0804 20:57:32.758735 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 157000: experiment/transformer/transformer_small/output/model.ckpt-157000
I0804 20:57:32.813354 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.68024
I0804 20:57:32.814395 140200711067520 basic_session_run_hooks.py:260] loss = 1.0443491, step = 157000 (59.515 sec)
I0804 20:57:36.093532 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4869
I0804 20:57:36.095075 140200711067520 basic_session_run_hooks.py:260] loss = 1.1270049, step = 157100 (3.281 sec)
I0804 20:57:39.334284 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8568
I0804 20:57:39.335658 140200711067520 basic_session_run_hooks.py:260] loss = 1.073313, step = 157200 (3.241 sec)
I0804 20:57:42.565376 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9493
I0804 20:57:42.566787 140200711067520 basic_session_run_hooks.py:260] loss = 1.0658575, step = 157300 (3.231 sec)
I0804 20:57:45.756904 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3329
I0804 20:57:45.758250 140200711067520 basic_session_run_hooks.py:260] loss = 1.101896, step = 157400 (3.191 sec)
I0804 20:57:48.943974 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3768
I0804 20:57:48.945234 140200711067520 basic_session_run_hooks.py:260] loss = 1.119426, step = 157500 (3.187 sec)
I0804 20:57:52.129685 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3902
I0804 20:57:52.131340 140200711067520 basic_session_run_hooks.py:260] loss = 1.1392746, step = 157600 (3.186 sec)
I0804 20:57:55.349040 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0621
I0804 20:57:55.350382 140200711067520 basic_session_run_hooks.py:260] loss = 1.0707831, step = 157700 (3.219 sec)
I0804 20:57:58.710245 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.7513
I0804 20:57:58.711761 140200711067520 basic_session_run_hooks.py:260] loss = 1.0990199, step = 157800 (3.361 sec)
I0804 20:58:01.947100 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8943
I0804 20:58:01.948683 140200711067520 basic_session_run_hooks.py:260] loss = 1.0427384, step = 157900 (3.237 sec)
I0804 20:58:05.130392 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 158000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:58:05.429639 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:58:05.469924 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3859
I0804 20:58:05.470871 140200711067520 basic_session_run_hooks.py:260] loss = 1.1038817, step = 158000 (3.522 sec)
I0804 20:58:08.675814 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1928
I0804 20:58:08.677170 140200711067520 basic_session_run_hooks.py:260] loss = 1.0486434, step = 158100 (3.206 sec)
I0804 20:58:11.886863 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1426
I0804 20:58:11.888392 140200711067520 basic_session_run_hooks.py:260] loss = 0.9865448, step = 158200 (3.211 sec)
I0804 20:58:15.082921 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2885
I0804 20:58:15.084486 140200711067520 basic_session_run_hooks.py:260] loss = 1.0629056, step = 158300 (3.196 sec)
I0804 20:58:18.288377 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1968
I0804 20:58:18.290014 140200711067520 basic_session_run_hooks.py:260] loss = 1.0987612, step = 158400 (3.206 sec)
I0804 20:58:21.496799 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1681
I0804 20:58:21.498003 140200711067520 basic_session_run_hooks.py:260] loss = 1.0042384, step = 158500 (3.208 sec)
I0804 20:58:24.720069 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0243
I0804 20:58:24.721518 140200711067520 basic_session_run_hooks.py:260] loss = 1.0845464, step = 158600 (3.224 sec)
I0804 20:58:27.921502 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2362
I0804 20:58:27.922884 140200711067520 basic_session_run_hooks.py:260] loss = 1.0295441, step = 158700 (3.201 sec)
I0804 20:58:31.131860 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1489
I0804 20:58:31.133266 140200711067520 basic_session_run_hooks.py:260] loss = 1.1830646, step = 158800 (3.210 sec)
I0804 20:58:34.332832 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2406
I0804 20:58:34.333919 140200711067520 basic_session_run_hooks.py:260] loss = 1.0845491, step = 158900 (3.201 sec)
I0804 20:58:37.518069 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 159000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:58:37.818718 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:58:37.863716 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3214
I0804 20:58:37.864881 140200711067520 basic_session_run_hooks.py:260] loss = 1.073263, step = 159000 (3.531 sec)
I0804 20:58:41.050088 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3839
I0804 20:58:41.051318 140200711067520 basic_session_run_hooks.py:260] loss = 1.0235358, step = 159100 (3.186 sec)
I0804 20:58:44.242548 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.324
I0804 20:58:44.244369 140200711067520 basic_session_run_hooks.py:260] loss = 1.0139016, step = 159200 (3.193 sec)
I0804 20:58:47.482202 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8673
I0804 20:58:47.483529 140200711067520 basic_session_run_hooks.py:260] loss = 1.0727465, step = 159300 (3.239 sec)
I0804 20:58:50.673264 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3377
I0804 20:58:50.674716 140200711067520 basic_session_run_hooks.py:260] loss = 1.0366877, step = 159400 (3.191 sec)
I0804 20:58:53.884790 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1377
I0804 20:58:53.886161 140200711067520 basic_session_run_hooks.py:260] loss = 1.055187, step = 159500 (3.211 sec)
I0804 20:58:57.067445 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4204
I0804 20:58:57.068689 140200711067520 basic_session_run_hooks.py:260] loss = 1.0260417, step = 159600 (3.183 sec)
I0804 20:59:00.281316 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1156
I0804 20:59:00.282775 140200711067520 basic_session_run_hooks.py:260] loss = 1.0039077, step = 159700 (3.214 sec)
I0804 20:59:03.509852 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9733
I0804 20:59:03.511274 140200711067520 basic_session_run_hooks.py:260] loss = 1.0613695, step = 159800 (3.229 sec)
I0804 20:59:06.735996 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9967
I0804 20:59:06.737281 140200711067520 basic_session_run_hooks.py:260] loss = 1.0326984, step = 159900 (3.226 sec)
I0804 20:59:09.914151 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 160000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:59:10.208259 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:59:10.244289 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5038
I0804 20:59:10.245533 140200711067520 basic_session_run_hooks.py:260] loss = 1.1549088, step = 160000 (3.508 sec)
I0804 20:59:13.470342 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9976
I0804 20:59:13.471781 140200711067520 basic_session_run_hooks.py:260] loss = 1.0209372, step = 160100 (3.226 sec)
I0804 20:59:16.663590 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.317
I0804 20:59:16.665004 140200711067520 basic_session_run_hooks.py:260] loss = 1.1096263, step = 160200 (3.193 sec)
I0804 20:59:19.881456 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.076
I0804 20:59:19.882578 140200711067520 basic_session_run_hooks.py:260] loss = 1.1822932, step = 160300 (3.218 sec)
I0804 20:59:23.091952 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1474
I0804 20:59:23.093139 140200711067520 basic_session_run_hooks.py:260] loss = 1.0539265, step = 160400 (3.211 sec)
I0804 20:59:26.309362 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0809
I0804 20:59:26.310740 140200711067520 basic_session_run_hooks.py:260] loss = 1.0767468, step = 160500 (3.218 sec)
I0804 20:59:29.488052 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4597
I0804 20:59:29.489346 140200711067520 basic_session_run_hooks.py:260] loss = 1.0746254, step = 160600 (3.179 sec)
I0804 20:59:32.672971 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3979
I0804 20:59:32.674013 140200711067520 basic_session_run_hooks.py:260] loss = 1.0412325, step = 160700 (3.185 sec)
I0804 20:59:35.886508 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1184
I0804 20:59:35.887633 140200711067520 basic_session_run_hooks.py:260] loss = 1.1293619, step = 160800 (3.214 sec)
I0804 20:59:39.111226 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0104
I0804 20:59:39.112941 140200711067520 basic_session_run_hooks.py:260] loss = 1.0621328, step = 160900 (3.225 sec)
I0804 20:59:42.316816 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 161000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 20:59:42.606099 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 20:59:42.641999 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3223
I0804 20:59:42.643307 140200711067520 basic_session_run_hooks.py:260] loss = 1.0547496, step = 161000 (3.530 sec)
I0804 20:59:45.863824 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0384
I0804 20:59:45.865090 140200711067520 basic_session_run_hooks.py:260] loss = 1.059844, step = 161100 (3.222 sec)
I0804 20:59:49.083394 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0602
I0804 20:59:49.084532 140200711067520 basic_session_run_hooks.py:260] loss = 1.0660617, step = 161200 (3.219 sec)
I0804 20:59:52.292702 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1598
I0804 20:59:52.293916 140200711067520 basic_session_run_hooks.py:260] loss = 1.0874394, step = 161300 (3.209 sec)
I0804 20:59:55.480351 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3711
I0804 20:59:55.481774 140200711067520 basic_session_run_hooks.py:260] loss = 1.0429276, step = 161400 (3.188 sec)
I0804 20:59:58.697150 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0864
I0804 20:59:58.698500 140200711067520 basic_session_run_hooks.py:260] loss = 1.0958437, step = 161500 (3.217 sec)
I0804 21:00:01.918036 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0479
I0804 21:00:01.919576 140200711067520 basic_session_run_hooks.py:260] loss = 1.0321612, step = 161600 (3.221 sec)
I0804 21:00:05.158549 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8588
I0804 21:00:05.160223 140200711067520 basic_session_run_hooks.py:260] loss = 1.1007314, step = 161700 (3.241 sec)
I0804 21:00:08.343948 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.393
I0804 21:00:08.345092 140200711067520 basic_session_run_hooks.py:260] loss = 1.120782, step = 161800 (3.185 sec)
I0804 21:00:11.558612 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1074
I0804 21:00:11.559917 140200711067520 basic_session_run_hooks.py:260] loss = 1.0449996, step = 161900 (3.215 sec)
I0804 21:00:14.699342 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 162000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:00:15.000782 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:00:15.038522 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7365
I0804 21:00:15.039640 140200711067520 basic_session_run_hooks.py:260] loss = 1.074646, step = 162000 (3.480 sec)
I0804 21:00:18.242626 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.21
I0804 21:00:18.243869 140200711067520 basic_session_run_hooks.py:260] loss = 1.1724113, step = 162100 (3.204 sec)
I0804 21:00:21.445054 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2261
I0804 21:00:21.446209 140200711067520 basic_session_run_hooks.py:260] loss = 1.1200558, step = 162200 (3.202 sec)
I0804 21:00:24.671802 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.991
I0804 21:00:24.672982 140200711067520 basic_session_run_hooks.py:260] loss = 1.0131927, step = 162300 (3.227 sec)
I0804 21:00:27.877397 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1954
I0804 21:00:27.878986 140200711067520 basic_session_run_hooks.py:260] loss = 1.0953544, step = 162400 (3.206 sec)
I0804 21:00:31.128850 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7556
I0804 21:00:31.130044 140200711067520 basic_session_run_hooks.py:260] loss = 1.0952698, step = 162500 (3.251 sec)
I0804 21:00:34.335189 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1881
I0804 21:00:34.336538 140200711067520 basic_session_run_hooks.py:260] loss = 1.111311, step = 162600 (3.206 sec)
I0804 21:00:37.527775 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3226
I0804 21:00:37.529122 140200711067520 basic_session_run_hooks.py:260] loss = 1.1794153, step = 162700 (3.193 sec)
I0804 21:00:40.754607 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9901
I0804 21:00:40.755802 140200711067520 basic_session_run_hooks.py:260] loss = 1.1108775, step = 162800 (3.227 sec)
I0804 21:00:43.969494 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1055
I0804 21:00:43.970876 140200711067520 basic_session_run_hooks.py:260] loss = 1.0105876, step = 162900 (3.215 sec)
I0804 21:00:47.161537 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 163000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:00:47.454278 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:00:47.496488 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3524
I0804 21:00:47.497650 140200711067520 basic_session_run_hooks.py:260] loss = 1.112472, step = 163000 (3.527 sec)
I0804 21:00:50.714590 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0747
I0804 21:00:50.715996 140200711067520 basic_session_run_hooks.py:260] loss = 1.0630313, step = 163100 (3.218 sec)
I0804 21:00:53.927289 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.131
I0804 21:00:53.933225 140200711067520 basic_session_run_hooks.py:260] loss = 1.1305287, step = 163200 (3.217 sec)
I0804 21:00:57.170376 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8302
I0804 21:00:57.171697 140200711067520 basic_session_run_hooks.py:260] loss = 1.0763088, step = 163300 (3.238 sec)
I0804 21:01:00.385262 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1052
I0804 21:01:00.386806 140200711067520 basic_session_run_hooks.py:260] loss = 1.0795463, step = 163400 (3.215 sec)
I0804 21:01:03.631774 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8022
I0804 21:01:03.632899 140200711067520 basic_session_run_hooks.py:260] loss = 1.0475544, step = 163500 (3.246 sec)
I0804 21:01:06.864266 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.936
I0804 21:01:06.865724 140200711067520 basic_session_run_hooks.py:260] loss = 1.0950929, step = 163600 (3.233 sec)
I0804 21:01:10.115347 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7589
I0804 21:01:10.116937 140200711067520 basic_session_run_hooks.py:260] loss = 1.1396297, step = 163700 (3.251 sec)
I0804 21:01:13.317168 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2323
I0804 21:01:13.318566 140200711067520 basic_session_run_hooks.py:260] loss = 1.1135358, step = 163800 (3.202 sec)
I0804 21:01:16.524599 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1778
I0804 21:01:16.525846 140200711067520 basic_session_run_hooks.py:260] loss = 1.0567364, step = 163900 (3.207 sec)
I0804 21:01:19.684749 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 164000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:01:19.969138 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:01:20.009719 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6931
I0804 21:01:20.010875 140200711067520 basic_session_run_hooks.py:260] loss = 1.0971614, step = 164000 (3.485 sec)
I0804 21:01:23.260919 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7581
I0804 21:01:23.262125 140200711067520 basic_session_run_hooks.py:260] loss = 0.9931682, step = 164100 (3.251 sec)
I0804 21:01:26.465263 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2076
I0804 21:01:26.466574 140200711067520 basic_session_run_hooks.py:260] loss = 1.0662081, step = 164200 (3.204 sec)
I0804 21:01:29.652025 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3798
I0804 21:01:29.653338 140200711067520 basic_session_run_hooks.py:260] loss = 1.0806568, step = 164300 (3.187 sec)
I0804 21:01:32.852636 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.244
I0804 21:01:32.853988 140200711067520 basic_session_run_hooks.py:260] loss = 1.0665902, step = 164400 (3.201 sec)
I0804 21:01:36.039121 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3828
I0804 21:01:36.040671 140200711067520 basic_session_run_hooks.py:260] loss = 1.0581583, step = 164500 (3.187 sec)
I0804 21:01:39.235824 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2821
I0804 21:01:39.237305 140200711067520 basic_session_run_hooks.py:260] loss = 1.0431126, step = 164600 (3.197 sec)
I0804 21:01:42.428965 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.317
I0804 21:01:42.430517 140200711067520 basic_session_run_hooks.py:260] loss = 1.015805, step = 164700 (3.193 sec)
I0804 21:01:45.636107 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1805
I0804 21:01:45.637690 140200711067520 basic_session_run_hooks.py:260] loss = 1.0693618, step = 164800 (3.207 sec)
I0804 21:01:48.829131 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3181
I0804 21:01:48.830480 140200711067520 basic_session_run_hooks.py:260] loss = 1.0932623, step = 164900 (3.193 sec)
I0804 21:01:51.952279 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 165000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:01:52.243875 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:01:52.287719 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.9135
I0804 21:01:52.288899 140200711067520 basic_session_run_hooks.py:260] loss = 1.0613803, step = 165000 (3.458 sec)
I0804 21:01:55.429398 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8304
I0804 21:01:55.430878 140200711067520 basic_session_run_hooks.py:260] loss = 1.0108758, step = 165100 (3.142 sec)
I0804 21:01:58.583474 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.705
I0804 21:01:58.584851 140200711067520 basic_session_run_hooks.py:260] loss = 1.1201148, step = 165200 (3.154 sec)
I0804 21:02:01.727272 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8085
I0804 21:02:01.728549 140200711067520 basic_session_run_hooks.py:260] loss = 1.1637936, step = 165300 (3.144 sec)
I0804 21:02:04.861276 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9081
I0804 21:02:04.862833 140200711067520 basic_session_run_hooks.py:260] loss = 1.002117, step = 165400 (3.134 sec)
I0804 21:02:08.020650 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6517
I0804 21:02:08.022023 140200711067520 basic_session_run_hooks.py:260] loss = 1.0501999, step = 165500 (3.159 sec)
I0804 21:02:11.173171 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7207
I0804 21:02:11.174358 140200711067520 basic_session_run_hooks.py:260] loss = 1.1588944, step = 165600 (3.152 sec)
I0804 21:02:14.393781 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.05
I0804 21:02:14.395173 140200711067520 basic_session_run_hooks.py:260] loss = 1.0646712, step = 165700 (3.221 sec)
I0804 21:02:17.596075 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2279
I0804 21:02:17.597627 140200711067520 basic_session_run_hooks.py:260] loss = 1.0907233, step = 165800 (3.202 sec)
I0804 21:02:20.786189 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3466
I0804 21:02:20.787619 140200711067520 basic_session_run_hooks.py:260] loss = 1.1208224, step = 165900 (3.190 sec)
I0804 21:02:23.947138 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 166000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:02:24.255335 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:02:24.302588 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.438
I0804 21:02:24.303714 140200711067520 basic_session_run_hooks.py:260] loss = 1.1209458, step = 166000 (3.516 sec)
I0804 21:02:27.517103 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1091
I0804 21:02:27.518347 140200711067520 basic_session_run_hooks.py:260] loss = 1.0499808, step = 166100 (3.215 sec)
I0804 21:02:30.708253 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3367
I0804 21:02:30.709539 140200711067520 basic_session_run_hooks.py:260] loss = 1.0265752, step = 166200 (3.191 sec)
I0804 21:02:33.911674 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2167
I0804 21:02:33.912965 140200711067520 basic_session_run_hooks.py:260] loss = 1.0603095, step = 166300 (3.203 sec)
I0804 21:02:37.116224 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2056
I0804 21:02:37.117614 140200711067520 basic_session_run_hooks.py:260] loss = 1.1372999, step = 166400 (3.205 sec)
I0804 21:02:40.275959 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6481
I0804 21:02:40.277263 140200711067520 basic_session_run_hooks.py:260] loss = 1.05218, step = 166500 (3.160 sec)
I0804 21:02:43.454805 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4581
I0804 21:02:43.456534 140200711067520 basic_session_run_hooks.py:260] loss = 1.1252898, step = 166600 (3.179 sec)
I0804 21:02:46.612650 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6669
I0804 21:02:46.613966 140200711067520 basic_session_run_hooks.py:260] loss = 1.050884, step = 166700 (3.157 sec)
I0804 21:02:49.773468 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6376
I0804 21:02:49.774794 140200711067520 basic_session_run_hooks.py:260] loss = 1.077137, step = 166800 (3.161 sec)
I0804 21:02:52.945494 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5255
I0804 21:02:52.946920 140200711067520 basic_session_run_hooks.py:260] loss = 1.0583732, step = 166900 (3.172 sec)
I0804 21:02:56.090473 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 167000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:02:56.382445 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:02:56.416150 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8127
I0804 21:02:56.417226 140200711067520 basic_session_run_hooks.py:260] loss = 1.0268842, step = 167000 (3.470 sec)
I0804 21:02:59.590251 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5052
I0804 21:02:59.591315 140200711067520 basic_session_run_hooks.py:260] loss = 1.1151906, step = 167100 (3.174 sec)
I0804 21:03:02.748465 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6636
I0804 21:03:02.749819 140200711067520 basic_session_run_hooks.py:260] loss = 1.0448648, step = 167200 (3.159 sec)
I0804 21:03:05.936754 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3646
I0804 21:03:05.938044 140200711067520 basic_session_run_hooks.py:260] loss = 1.0450319, step = 167300 (3.188 sec)
I0804 21:03:09.135768 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2596
I0804 21:03:09.137174 140200711067520 basic_session_run_hooks.py:260] loss = 1.0191898, step = 167400 (3.199 sec)
I0804 21:03:12.312022 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4837
I0804 21:03:12.313217 140200711067520 basic_session_run_hooks.py:260] loss = 1.047217, step = 167500 (3.176 sec)
I0804 21:03:15.464163 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7245
I0804 21:03:15.465577 140200711067520 basic_session_run_hooks.py:260] loss = 1.0748862, step = 167600 (3.152 sec)
I0804 21:03:18.799638 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.9807
I0804 21:03:18.800999 140200711067520 basic_session_run_hooks.py:260] loss = 1.058603, step = 167700 (3.335 sec)
I0804 21:03:22.018219 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0696
I0804 21:03:22.019457 140200711067520 basic_session_run_hooks.py:260] loss = 1.067946, step = 167800 (3.218 sec)
I0804 21:03:25.238476 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0537
I0804 21:03:25.240005 140200711067520 basic_session_run_hooks.py:260] loss = 1.1107543, step = 167900 (3.221 sec)
I0804 21:03:28.407906 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 168000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:03:28.712273 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:03:28.753982 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4452
I0804 21:03:28.754941 140200711067520 basic_session_run_hooks.py:260] loss = 1.0913178, step = 168000 (3.515 sec)
I0804 21:03:31.971705 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.078
I0804 21:03:31.973282 140200711067520 basic_session_run_hooks.py:260] loss = 1.0671248, step = 168100 (3.218 sec)
I0804 21:03:35.179383 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.175
I0804 21:03:35.180749 140200711067520 basic_session_run_hooks.py:260] loss = 1.0643165, step = 168200 (3.207 sec)
I0804 21:03:38.378551 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2583
I0804 21:03:38.379702 140200711067520 basic_session_run_hooks.py:260] loss = 1.0466758, step = 168300 (3.199 sec)
I0804 21:03:41.583551 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2013
I0804 21:03:41.585072 140200711067520 basic_session_run_hooks.py:260] loss = 1.0007777, step = 168400 (3.205 sec)
I0804 21:03:44.783250 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2527
I0804 21:03:44.784816 140200711067520 basic_session_run_hooks.py:260] loss = 1.0953603, step = 168500 (3.200 sec)
I0804 21:03:47.987109 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2123
I0804 21:03:47.988662 140200711067520 basic_session_run_hooks.py:260] loss = 0.9154692, step = 168600 (3.204 sec)
I0804 21:03:51.202494 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1007
I0804 21:03:51.203768 140200711067520 basic_session_run_hooks.py:260] loss = 1.0041019, step = 168700 (3.215 sec)
I0804 21:03:54.374614 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5244
I0804 21:03:54.375956 140200711067520 basic_session_run_hooks.py:260] loss = 1.1102176, step = 168800 (3.172 sec)
I0804 21:03:57.574693 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2493
I0804 21:03:57.575991 140200711067520 basic_session_run_hooks.py:260] loss = 1.1485691, step = 168900 (3.200 sec)
I0804 21:04:00.725996 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 169000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:04:01.020580 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:04:01.055402 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7295
I0804 21:04:01.056523 140200711067520 basic_session_run_hooks.py:260] loss = 1.1135081, step = 169000 (3.481 sec)
I0804 21:04:04.240078 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4007
I0804 21:04:04.241521 140200711067520 basic_session_run_hooks.py:260] loss = 0.9956322, step = 169100 (3.185 sec)
I0804 21:04:07.471778 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9434
I0804 21:04:07.472974 140200711067520 basic_session_run_hooks.py:260] loss = 1.1419022, step = 169200 (3.231 sec)
I0804 21:04:10.652741 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4371
I0804 21:04:10.653944 140200711067520 basic_session_run_hooks.py:260] loss = 1.0677744, step = 169300 (3.181 sec)
I0804 21:04:13.868829 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0935
I0804 21:04:13.870164 140200711067520 basic_session_run_hooks.py:260] loss = 1.0405437, step = 169400 (3.216 sec)
I0804 21:04:17.042262 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5117
I0804 21:04:17.043932 140200711067520 basic_session_run_hooks.py:260] loss = 1.0330676, step = 169500 (3.174 sec)
I0804 21:04:20.227981 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3901
I0804 21:04:20.229163 140200711067520 basic_session_run_hooks.py:260] loss = 1.1065995, step = 169600 (3.185 sec)
I0804 21:04:23.445677 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0783
I0804 21:04:23.447131 140200711067520 basic_session_run_hooks.py:260] loss = 1.0538632, step = 169700 (3.218 sec)
I0804 21:04:26.656294 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1465
I0804 21:04:26.657855 140200711067520 basic_session_run_hooks.py:260] loss = 1.089064, step = 169800 (3.211 sec)
I0804 21:04:29.882941 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9919
I0804 21:04:29.884461 140200711067520 basic_session_run_hooks.py:260] loss = 1.0388798, step = 169900 (3.227 sec)
I0804 21:04:33.079351 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 170000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:04:33.371485 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:04:33.411826 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3376
I0804 21:04:33.412957 140200711067520 basic_session_run_hooks.py:260] loss = 1.178629, step = 170000 (3.529 sec)
I0804 21:04:36.621127 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1594
I0804 21:04:36.622502 140200711067520 basic_session_run_hooks.py:260] loss = 1.0795245, step = 170100 (3.210 sec)
I0804 21:04:39.826551 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1974
I0804 21:04:39.827801 140200711067520 basic_session_run_hooks.py:260] loss = 1.0510434, step = 170200 (3.205 sec)
I0804 21:04:43.015314 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3599
I0804 21:04:43.017000 140200711067520 basic_session_run_hooks.py:260] loss = 0.9508452, step = 170300 (3.189 sec)
I0804 21:04:46.214909 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2539
I0804 21:04:46.216360 140200711067520 basic_session_run_hooks.py:260] loss = 1.0459921, step = 170400 (3.199 sec)
I0804 21:04:49.407620 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3214
I0804 21:04:49.408990 140200711067520 basic_session_run_hooks.py:260] loss = 1.0559808, step = 170500 (3.193 sec)
I0804 21:04:52.603057 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2946
I0804 21:04:52.604330 140200711067520 basic_session_run_hooks.py:260] loss = 1.1150645, step = 170600 (3.195 sec)
I0804 21:04:55.822006 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0661
I0804 21:04:55.823186 140200711067520 basic_session_run_hooks.py:260] loss = 1.0376198, step = 170700 (3.219 sec)
I0804 21:04:59.039823 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.077
I0804 21:04:59.041478 140200711067520 basic_session_run_hooks.py:260] loss = 1.0255818, step = 170800 (3.218 sec)
I0804 21:05:02.242566 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2234
I0804 21:05:02.244074 140200711067520 basic_session_run_hooks.py:260] loss = 1.084767, step = 170900 (3.203 sec)
I0804 21:05:05.410229 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 171000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:05:05.704015 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:05:05.741399 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5807
I0804 21:05:05.742714 140200711067520 basic_session_run_hooks.py:260] loss = 1.0317827, step = 171000 (3.499 sec)
I0804 21:05:08.921651 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4441
I0804 21:05:08.922944 140200711067520 basic_session_run_hooks.py:260] loss = 1.0811278, step = 171100 (3.180 sec)
I0804 21:05:12.102594 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4375
I0804 21:05:12.103914 140200711067520 basic_session_run_hooks.py:260] loss = 0.9747795, step = 171200 (3.181 sec)
I0804 21:05:15.302529 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2506
I0804 21:05:15.303926 140200711067520 basic_session_run_hooks.py:260] loss = 1.0984854, step = 171300 (3.200 sec)
I0804 21:05:18.490727 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3654
I0804 21:05:18.492232 140200711067520 basic_session_run_hooks.py:260] loss = 1.1006233, step = 171400 (3.188 sec)
I0804 21:05:21.661401 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5391
I0804 21:05:21.663009 140200711067520 basic_session_run_hooks.py:260] loss = 1.0900131, step = 171500 (3.171 sec)
I0804 21:05:24.869681 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1695
I0804 21:05:24.871146 140200711067520 basic_session_run_hooks.py:260] loss = 1.0651176, step = 171600 (3.208 sec)
I0804 21:05:28.051258 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4308
I0804 21:05:28.052651 140200711067520 basic_session_run_hooks.py:260] loss = 1.1108102, step = 171700 (3.182 sec)
I0804 21:05:31.216687 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5913
I0804 21:05:31.218245 140200711067520 basic_session_run_hooks.py:260] loss = 1.0400906, step = 171800 (3.166 sec)
I0804 21:05:34.372334 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6892
I0804 21:05:34.373536 140200711067520 basic_session_run_hooks.py:260] loss = 1.013728, step = 171900 (3.155 sec)
I0804 21:05:37.516016 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 172000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:05:37.816596 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:05:37.858522 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6847
I0804 21:05:37.859537 140200711067520 basic_session_run_hooks.py:260] loss = 1.1439543, step = 172000 (3.486 sec)
I0804 21:05:41.037403 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4576
I0804 21:05:41.038753 140200711067520 basic_session_run_hooks.py:260] loss = 1.0280756, step = 172100 (3.179 sec)
I0804 21:05:44.226022 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3615
I0804 21:05:44.227346 140200711067520 basic_session_run_hooks.py:260] loss = 1.015517, step = 172200 (3.189 sec)
I0804 21:05:47.421156 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2976
I0804 21:05:47.422645 140200711067520 basic_session_run_hooks.py:260] loss = 1.0972959, step = 172300 (3.195 sec)
I0804 21:05:50.670453 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7759
I0804 21:05:50.671945 140200711067520 basic_session_run_hooks.py:260] loss = 1.0525677, step = 172400 (3.249 sec)
I0804 21:05:53.881539 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1425
I0804 21:05:53.882757 140200711067520 basic_session_run_hooks.py:260] loss = 1.1206003, step = 172500 (3.211 sec)
I0804 21:05:57.116128 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9155
I0804 21:05:57.117254 140200711067520 basic_session_run_hooks.py:260] loss = 1.0392224, step = 172600 (3.234 sec)
I0804 21:06:00.338889 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0291
I0804 21:06:00.340344 140200711067520 basic_session_run_hooks.py:260] loss = 1.0326785, step = 172700 (3.223 sec)
I0804 21:06:03.557292 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0713
I0804 21:06:03.558717 140200711067520 basic_session_run_hooks.py:260] loss = 1.121066, step = 172800 (3.218 sec)
I0804 21:06:06.800390 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8347
I0804 21:06:06.801958 140200711067520 basic_session_run_hooks.py:260] loss = 1.087784, step = 172900 (3.243 sec)
I0804 21:06:09.987477 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 173000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:06:10.276737 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:06:10.318856 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4213
I0804 21:06:10.320149 140200711067520 basic_session_run_hooks.py:260] loss = 1.097346, step = 173000 (3.518 sec)
I0804 21:06:13.605262 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4287
I0804 21:06:13.606986 140200711067520 basic_session_run_hooks.py:260] loss = 1.0556792, step = 173100 (3.287 sec)
I0804 21:06:16.861186 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7132
I0804 21:06:16.862283 140200711067520 basic_session_run_hooks.py:260] loss = 0.95227927, step = 173200 (3.255 sec)
I0804 21:06:20.068829 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1754
I0804 21:06:20.069952 140200711067520 basic_session_run_hooks.py:260] loss = 1.0570022, step = 173300 (3.208 sec)
I0804 21:06:23.268121 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2569
I0804 21:06:23.269459 140200711067520 basic_session_run_hooks.py:260] loss = 1.1068089, step = 173400 (3.199 sec)
I0804 21:06:26.442936 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.498
I0804 21:06:26.444454 140200711067520 basic_session_run_hooks.py:260] loss = 0.97485566, step = 173500 (3.175 sec)
I0804 21:06:29.617959 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4959
I0804 21:06:29.619304 140200711067520 basic_session_run_hooks.py:260] loss = 1.0965952, step = 173600 (3.175 sec)
I0804 21:06:32.768971 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7358
I0804 21:06:32.770068 140200711067520 basic_session_run_hooks.py:260] loss = 1.055892, step = 173700 (3.151 sec)
I0804 21:06:35.938613 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5493
I0804 21:06:35.940063 140200711067520 basic_session_run_hooks.py:260] loss = 1.115946, step = 173800 (3.170 sec)
I0804 21:06:39.127319 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3607
I0804 21:06:39.128726 140200711067520 basic_session_run_hooks.py:260] loss = 1.079741, step = 173900 (3.189 sec)
I0804 21:06:42.301584 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 174000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:06:42.610078 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 21:06:42.611384 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 21:06:42.756519 140200711067520 estimator.py:1145] Calling model_fn.
I0804 21:06:42.757540 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 21:06:42.757934 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 21:06:42.758026 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 21:06:42.758106 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 21:06:42.758176 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 21:06:42.758259 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 21:06:42.758325 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 21:06:43.107043 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 21:06:43.168584 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 21:06:43.309158 140200711067520 t2t_model.py:2172] Building model body
I0804 21:06:43.998783 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 21:06:44.908999 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 21:06:44.927495 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T21:06:44Z
I0804 21:06:45.090164 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 21:06:45.090765: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:06:45.091163: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 21:06:45.091255: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 21:06:45.091282: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 21:06:45.091304: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 21:06:45.091330: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 21:06:45.091351: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 21:06:45.091370: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 21:06:45.091390: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 21:06:45.091525: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:06:45.091930: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:06:45.092249: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 21:06:45.092292: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 21:06:45.092305: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 21:06:45.092315: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 21:06:45.092610: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:06:45.092998: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:06:45.093331: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 21:06:45.094797 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-174000
I0804 21:06:45.310090 140200711067520 session_manager.py:500] Running local_init_op.
I0804 21:06:45.359185 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 21:06:51.377408 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 21:06:56.722464 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 21:07:02.061596 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 21:07:07.440887 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 21:07:12.773241 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 21:07:18.086076 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 21:07:23.377994 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 21:07:28.716120 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 21:07:34.024100 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 21:07:38.874696 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-21:07:38
I0804 21:07:38.874920 140200711067520 estimator.py:2039] Saving dict for global step 174000: global_step = 174000, loss = 1.166916, metrics-paper_generation_problem/targets/accuracy = 0.67607594, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.88593876, metrics-paper_generation_problem/targets/approx_bleu_score = 0.4913871, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.166952, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.584894, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.69678706
I0804 21:07:38.875390 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 174000: experiment/transformer/transformer_small/output/model.ckpt-174000
I0804 21:07:38.930725 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.67214
I0804 21:07:38.931838 140200711067520 basic_session_run_hooks.py:260] loss = 1.0552714, step = 174000 (59.803 sec)
I0804 21:07:42.163580 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9331
I0804 21:07:42.165004 140200711067520 basic_session_run_hooks.py:260] loss = 1.016604, step = 174100 (3.233 sec)
I0804 21:07:45.350844 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3748
I0804 21:07:45.352299 140200711067520 basic_session_run_hooks.py:260] loss = 1.1163857, step = 174200 (3.187 sec)
I0804 21:07:48.528046 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4741
I0804 21:07:48.529439 140200711067520 basic_session_run_hooks.py:260] loss = 1.0160387, step = 174300 (3.177 sec)
I0804 21:07:51.724339 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2864
I0804 21:07:51.725711 140200711067520 basic_session_run_hooks.py:260] loss = 1.031689, step = 174400 (3.196 sec)
I0804 21:07:54.954302 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9601
I0804 21:07:54.955561 140200711067520 basic_session_run_hooks.py:260] loss = 1.1445624, step = 174500 (3.230 sec)
I0804 21:07:58.171653 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0814
I0804 21:07:58.172995 140200711067520 basic_session_run_hooks.py:260] loss = 1.1110963, step = 174600 (3.217 sec)
I0804 21:08:01.420485 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7821
I0804 21:08:01.422062 140200711067520 basic_session_run_hooks.py:260] loss = 1.0943941, step = 174700 (3.249 sec)
I0804 21:08:04.678104 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6955
I0804 21:08:04.679706 140200711067520 basic_session_run_hooks.py:260] loss = 1.0409884, step = 174800 (3.258 sec)
I0804 21:08:07.891046 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1243
I0804 21:08:07.892654 140200711067520 basic_session_run_hooks.py:260] loss = 1.0570114, step = 174900 (3.213 sec)
I0804 21:08:11.076216 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 175000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:08:11.379180 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:08:11.415683 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3715
I0804 21:08:11.417023 140200711067520 basic_session_run_hooks.py:260] loss = 1.0449759, step = 175000 (3.524 sec)
I0804 21:08:14.610809 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2978
I0804 21:08:14.612258 140200711067520 basic_session_run_hooks.py:260] loss = 1.0366834, step = 175100 (3.195 sec)
I0804 21:08:17.767905 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6747
I0804 21:08:17.769000 140200711067520 basic_session_run_hooks.py:260] loss = 1.0239378, step = 175200 (3.157 sec)
I0804 21:08:20.946554 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4602
I0804 21:08:20.947920 140200711067520 basic_session_run_hooks.py:260] loss = 0.99588364, step = 175300 (3.179 sec)
I0804 21:08:24.156116 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1566
I0804 21:08:24.157840 140200711067520 basic_session_run_hooks.py:260] loss = 1.0610142, step = 175400 (3.210 sec)
I0804 21:08:27.354071 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2701
I0804 21:08:27.355229 140200711067520 basic_session_run_hooks.py:260] loss = 1.0963137, step = 175500 (3.197 sec)
I0804 21:08:30.603532 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7745
I0804 21:08:30.604945 140200711067520 basic_session_run_hooks.py:260] loss = 1.0667025, step = 175600 (3.250 sec)
I0804 21:08:33.853128 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7728
I0804 21:08:33.854465 140200711067520 basic_session_run_hooks.py:260] loss = 1.0756187, step = 175700 (3.250 sec)
I0804 21:08:37.081162 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9784
I0804 21:08:37.082308 140200711067520 basic_session_run_hooks.py:260] loss = 0.99660504, step = 175800 (3.228 sec)
I0804 21:08:40.320312 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8724
I0804 21:08:40.321739 140200711067520 basic_session_run_hooks.py:260] loss = 1.0123423, step = 175900 (3.239 sec)
I0804 21:08:43.501462 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 176000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:08:43.790533 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:08:43.836779 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4374
I0804 21:08:43.837831 140200711067520 basic_session_run_hooks.py:260] loss = 0.9476761, step = 176000 (3.516 sec)
I0804 21:08:47.071665 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9134
I0804 21:08:47.073017 140200711067520 basic_session_run_hooks.py:260] loss = 1.1612628, step = 176100 (3.235 sec)
I0804 21:08:50.301263 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9636
I0804 21:08:50.302706 140200711067520 basic_session_run_hooks.py:260] loss = 1.114299, step = 176200 (3.230 sec)
I0804 21:08:53.522871 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0402
I0804 21:08:53.524053 140200711067520 basic_session_run_hooks.py:260] loss = 1.1296803, step = 176300 (3.221 sec)
I0804 21:08:56.702651 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4488
I0804 21:08:56.704058 140200711067520 basic_session_run_hooks.py:260] loss = 1.0648956, step = 176400 (3.180 sec)
I0804 21:08:59.857178 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7004
I0804 21:08:59.858883 140200711067520 basic_session_run_hooks.py:260] loss = 1.0516195, step = 176500 (3.155 sec)
I0804 21:09:03.002828 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7898
I0804 21:09:03.004095 140200711067520 basic_session_run_hooks.py:260] loss = 1.0424854, step = 176600 (3.145 sec)
I0804 21:09:06.177686 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4976
I0804 21:09:06.179125 140200711067520 basic_session_run_hooks.py:260] loss = 1.090082, step = 176700 (3.175 sec)
I0804 21:09:09.352906 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4938
I0804 21:09:09.354398 140200711067520 basic_session_run_hooks.py:260] loss = 1.0405405, step = 176800 (3.175 sec)
I0804 21:09:12.498063 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7951
I0804 21:09:12.499461 140200711067520 basic_session_run_hooks.py:260] loss = 1.1089559, step = 176900 (3.145 sec)
I0804 21:09:15.627707 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 177000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:09:15.927207 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:09:15.971643 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7884
I0804 21:09:15.972735 140200711067520 basic_session_run_hooks.py:260] loss = 1.0719739, step = 177000 (3.473 sec)
I0804 21:09:19.134298 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6193
I0804 21:09:19.135708 140200711067520 basic_session_run_hooks.py:260] loss = 1.015743, step = 177100 (3.163 sec)
I0804 21:09:22.355491 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0445
I0804 21:09:22.356783 140200711067520 basic_session_run_hooks.py:260] loss = 1.0379999, step = 177200 (3.221 sec)
I0804 21:09:25.604574 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.778
I0804 21:09:25.605686 140200711067520 basic_session_run_hooks.py:260] loss = 1.0788432, step = 177300 (3.249 sec)
I0804 21:09:28.847323 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8377
I0804 21:09:28.848518 140200711067520 basic_session_run_hooks.py:260] loss = 1.0742351, step = 177400 (3.243 sec)
I0804 21:09:32.065463 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0744
I0804 21:09:32.066703 140200711067520 basic_session_run_hooks.py:260] loss = 1.0445911, step = 177500 (3.218 sec)
I0804 21:09:35.400105 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.9878
I0804 21:09:35.401467 140200711067520 basic_session_run_hooks.py:260] loss = 1.0596051, step = 177600 (3.335 sec)
I0804 21:09:38.610807 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1458
I0804 21:09:38.612097 140200711067520 basic_session_run_hooks.py:260] loss = 1.0883538, step = 177700 (3.211 sec)
I0804 21:09:41.814340 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2155
I0804 21:09:41.815748 140200711067520 basic_session_run_hooks.py:260] loss = 1.0490352, step = 177800 (3.204 sec)
I0804 21:09:45.049080 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9146
I0804 21:09:45.050521 140200711067520 basic_session_run_hooks.py:260] loss = 1.0300976, step = 177900 (3.235 sec)
I0804 21:09:48.249240 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 178000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:09:48.557056 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:09:48.599881 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.1624
I0804 21:09:48.600881 140200711067520 basic_session_run_hooks.py:260] loss = 1.0834858, step = 178000 (3.550 sec)
I0804 21:09:51.808355 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1677
I0804 21:09:51.810104 140200711067520 basic_session_run_hooks.py:260] loss = 1.0399202, step = 178100 (3.209 sec)
I0804 21:09:55.006203 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2711
I0804 21:09:55.007527 140200711067520 basic_session_run_hooks.py:260] loss = 0.9333032, step = 178200 (3.197 sec)
I0804 21:09:58.222709 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0896
I0804 21:09:58.223966 140200711067520 basic_session_run_hooks.py:260] loss = 0.9664612, step = 178300 (3.216 sec)
I0804 21:10:01.431921 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1603
I0804 21:10:01.432975 140200711067520 basic_session_run_hooks.py:260] loss = 1.035809, step = 178400 (3.209 sec)
I0804 21:10:04.633787 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2318
I0804 21:10:04.634873 140200711067520 basic_session_run_hooks.py:260] loss = 0.9635471, step = 178500 (3.202 sec)
I0804 21:10:07.856103 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0336
I0804 21:10:07.857297 140200711067520 basic_session_run_hooks.py:260] loss = 1.0857157, step = 178600 (3.222 sec)
I0804 21:10:11.080060 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0179
I0804 21:10:11.081454 140200711067520 basic_session_run_hooks.py:260] loss = 1.0477518, step = 178700 (3.224 sec)
I0804 21:10:14.302520 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0322
I0804 21:10:14.304035 140200711067520 basic_session_run_hooks.py:260] loss = 1.0727117, step = 178800 (3.223 sec)
I0804 21:10:17.525293 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0291
I0804 21:10:17.526801 140200711067520 basic_session_run_hooks.py:260] loss = 1.1473124, step = 178900 (3.223 sec)
I0804 21:10:20.718652 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 179000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:10:21.014255 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:10:21.057328 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3119
I0804 21:10:21.058409 140200711067520 basic_session_run_hooks.py:260] loss = 1.1336948, step = 179000 (3.532 sec)
I0804 21:10:24.285679 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9758
I0804 21:10:24.287128 140200711067520 basic_session_run_hooks.py:260] loss = 1.0884743, step = 179100 (3.229 sec)
I0804 21:10:27.467037 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4332
I0804 21:10:27.468524 140200711067520 basic_session_run_hooks.py:260] loss = 0.986739, step = 179200 (3.181 sec)
I0804 21:10:30.649557 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4217
I0804 21:10:30.650829 140200711067520 basic_session_run_hooks.py:260] loss = 1.0789889, step = 179300 (3.182 sec)
I0804 21:10:33.833106 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4138
I0804 21:10:33.834541 140200711067520 basic_session_run_hooks.py:260] loss = 1.1239738, step = 179400 (3.184 sec)
I0804 21:10:37.034657 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2324
I0804 21:10:37.036262 140200711067520 basic_session_run_hooks.py:260] loss = 1.1281435, step = 179500 (3.202 sec)
I0804 21:10:40.215223 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4409
I0804 21:10:40.216456 140200711067520 basic_session_run_hooks.py:260] loss = 1.0682614, step = 179600 (3.180 sec)
I0804 21:10:43.409882 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3023
I0804 21:10:43.411279 140200711067520 basic_session_run_hooks.py:260] loss = 1.1679885, step = 179700 (3.195 sec)
I0804 21:10:46.600912 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3378
I0804 21:10:46.602267 140200711067520 basic_session_run_hooks.py:260] loss = 0.9820263, step = 179800 (3.191 sec)
I0804 21:10:49.814082 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1221
I0804 21:10:49.815570 140200711067520 basic_session_run_hooks.py:260] loss = 1.1709638, step = 179900 (3.213 sec)
I0804 21:10:52.975078 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 180000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:10:53.279802 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:10:53.318321 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5365
I0804 21:10:53.319783 140200711067520 basic_session_run_hooks.py:260] loss = 1.0327605, step = 180000 (3.504 sec)
I0804 21:10:56.548867 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9548
I0804 21:10:56.550151 140200711067520 basic_session_run_hooks.py:260] loss = 1.083517, step = 180100 (3.230 sec)
I0804 21:10:59.730084 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4346
I0804 21:10:59.731369 140200711067520 basic_session_run_hooks.py:260] loss = 1.0424366, step = 180200 (3.181 sec)
I0804 21:11:02.914627 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4016
I0804 21:11:02.915867 140200711067520 basic_session_run_hooks.py:260] loss = 1.0506784, step = 180300 (3.184 sec)
I0804 21:11:06.134640 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0581
I0804 21:11:06.136067 140200711067520 basic_session_run_hooks.py:260] loss = 1.0093179, step = 180400 (3.220 sec)
I0804 21:11:09.326445 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.328
I0804 21:11:09.327848 140200711067520 basic_session_run_hooks.py:260] loss = 1.0368416, step = 180500 (3.192 sec)
I0804 21:11:12.541770 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1009
I0804 21:11:12.543392 140200711067520 basic_session_run_hooks.py:260] loss = 1.0873386, step = 180600 (3.216 sec)
I0804 21:11:15.782890 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8536
I0804 21:11:15.784221 140200711067520 basic_session_run_hooks.py:260] loss = 1.0748937, step = 180700 (3.241 sec)
I0804 21:11:19.003278 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.052
I0804 21:11:19.004708 140200711067520 basic_session_run_hooks.py:260] loss = 1.1222411, step = 180800 (3.220 sec)
I0804 21:11:22.252585 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7761
I0804 21:11:22.254118 140200711067520 basic_session_run_hooks.py:260] loss = 1.0290138, step = 180900 (3.249 sec)
I0804 21:11:25.481243 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 181000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:11:25.780514 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:11:25.828388 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 27.9657
I0804 21:11:25.829500 140200711067520 basic_session_run_hooks.py:260] loss = 1.0697246, step = 181000 (3.575 sec)
I0804 21:11:29.071686 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8327
I0804 21:11:29.072909 140200711067520 basic_session_run_hooks.py:260] loss = 0.987813, step = 181100 (3.243 sec)
I0804 21:11:32.307577 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9035
I0804 21:11:32.308845 140200711067520 basic_session_run_hooks.py:260] loss = 1.0602225, step = 181200 (3.236 sec)
I0804 21:11:35.513651 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1907
I0804 21:11:35.514793 140200711067520 basic_session_run_hooks.py:260] loss = 1.1266702, step = 181300 (3.206 sec)
I0804 21:11:38.715125 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2356
I0804 21:11:38.716466 140200711067520 basic_session_run_hooks.py:260] loss = 1.0808277, step = 181400 (3.202 sec)
I0804 21:11:41.929196 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1131
I0804 21:11:41.930589 140200711067520 basic_session_run_hooks.py:260] loss = 1.0219295, step = 181500 (3.214 sec)
I0804 21:11:45.074025 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7982
I0804 21:11:45.075537 140200711067520 basic_session_run_hooks.py:260] loss = 1.0957868, step = 181600 (3.145 sec)
I0804 21:11:48.206454 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9243
I0804 21:11:48.207547 140200711067520 basic_session_run_hooks.py:260] loss = 1.0796834, step = 181700 (3.132 sec)
I0804 21:11:51.335309 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9604
I0804 21:11:51.336648 140200711067520 basic_session_run_hooks.py:260] loss = 1.1379254, step = 181800 (3.129 sec)
I0804 21:11:54.466641 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9354
I0804 21:11:54.468168 140200711067520 basic_session_run_hooks.py:260] loss = 1.0320176, step = 181900 (3.132 sec)
I0804 21:11:57.604089 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 182000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:11:57.902415 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:11:57.939308 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.796
I0804 21:11:57.940450 140200711067520 basic_session_run_hooks.py:260] loss = 1.0263377, step = 182000 (3.472 sec)
I0804 21:12:01.090367 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7357
I0804 21:12:01.091791 140200711067520 basic_session_run_hooks.py:260] loss = 1.0814099, step = 182100 (3.151 sec)
I0804 21:12:04.199465 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.1639
I0804 21:12:04.200930 140200711067520 basic_session_run_hooks.py:260] loss = 1.0398699, step = 182200 (3.109 sec)
I0804 21:12:07.377291 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4678
I0804 21:12:07.378705 140200711067520 basic_session_run_hooks.py:260] loss = 1.0688614, step = 182300 (3.178 sec)
I0804 21:12:10.585547 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1699
I0804 21:12:10.586723 140200711067520 basic_session_run_hooks.py:260] loss = 1.0969173, step = 182400 (3.208 sec)
I0804 21:12:13.786379 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2417
I0804 21:12:13.787710 140200711067520 basic_session_run_hooks.py:260] loss = 1.1036509, step = 182500 (3.201 sec)
I0804 21:12:16.997289 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1438
I0804 21:12:16.998618 140200711067520 basic_session_run_hooks.py:260] loss = 1.061188, step = 182600 (3.211 sec)
I0804 21:12:20.208121 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1445
I0804 21:12:20.209614 140200711067520 basic_session_run_hooks.py:260] loss = 1.0203865, step = 182700 (3.211 sec)
I0804 21:12:23.430697 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0311
I0804 21:12:23.432068 140200711067520 basic_session_run_hooks.py:260] loss = 1.0616156, step = 182800 (3.222 sec)
I0804 21:12:26.660789 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9589
I0804 21:12:26.662087 140200711067520 basic_session_run_hooks.py:260] loss = 1.0546091, step = 182900 (3.230 sec)
I0804 21:12:29.835552 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 183000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:12:30.134650 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:12:30.173165 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4706
I0804 21:12:30.174264 140200711067520 basic_session_run_hooks.py:260] loss = 1.0704308, step = 183000 (3.512 sec)
I0804 21:12:33.440168 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6093
I0804 21:12:33.441545 140200711067520 basic_session_run_hooks.py:260] loss = 1.0068833, step = 183100 (3.267 sec)
I0804 21:12:36.663322 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0256
I0804 21:12:36.664752 140200711067520 basic_session_run_hooks.py:260] loss = 1.0632727, step = 183200 (3.223 sec)
I0804 21:12:39.847099 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4092
I0804 21:12:39.848366 140200711067520 basic_session_run_hooks.py:260] loss = 1.129179, step = 183300 (3.184 sec)
I0804 21:12:43.062334 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1019
I0804 21:12:43.063775 140200711067520 basic_session_run_hooks.py:260] loss = 1.120728, step = 183400 (3.215 sec)
I0804 21:12:46.277245 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1052
I0804 21:12:46.278533 140200711067520 basic_session_run_hooks.py:260] loss = 1.0479198, step = 183500 (3.215 sec)
I0804 21:12:49.486033 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1642
I0804 21:12:49.487324 140200711067520 basic_session_run_hooks.py:260] loss = 1.0478634, step = 183600 (3.209 sec)
I0804 21:12:52.675384 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3543
I0804 21:12:52.676798 140200711067520 basic_session_run_hooks.py:260] loss = 1.1137748, step = 183700 (3.189 sec)
I0804 21:12:55.899513 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0164
I0804 21:12:55.900903 140200711067520 basic_session_run_hooks.py:260] loss = 1.1012747, step = 183800 (3.224 sec)
I0804 21:12:59.141345 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8467
I0804 21:12:59.142729 140200711067520 basic_session_run_hooks.py:260] loss = 1.0301468, step = 183900 (3.242 sec)
I0804 21:13:02.290358 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 184000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:13:02.595872 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:13:02.633832 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6326
I0804 21:13:02.635038 140200711067520 basic_session_run_hooks.py:260] loss = 1.0151076, step = 184000 (3.492 sec)
I0804 21:13:05.832191 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2664
I0804 21:13:05.833328 140200711067520 basic_session_run_hooks.py:260] loss = 1.0450017, step = 184100 (3.198 sec)
I0804 21:13:09.009923 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4689
I0804 21:13:09.011315 140200711067520 basic_session_run_hooks.py:260] loss = 1.1013525, step = 184200 (3.178 sec)
I0804 21:13:12.193442 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.412
I0804 21:13:12.194569 140200711067520 basic_session_run_hooks.py:260] loss = 1.0845447, step = 184300 (3.183 sec)
I0804 21:13:15.387898 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3041
I0804 21:13:15.389294 140200711067520 basic_session_run_hooks.py:260] loss = 1.058339, step = 184400 (3.195 sec)
I0804 21:13:18.567695 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4484
I0804 21:13:18.568835 140200711067520 basic_session_run_hooks.py:260] loss = 1.0837723, step = 184500 (3.180 sec)
I0804 21:13:21.758872 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3366
I0804 21:13:21.760132 140200711067520 basic_session_run_hooks.py:260] loss = 1.0357593, step = 184600 (3.191 sec)
I0804 21:13:25.020248 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6618
I0804 21:13:25.021935 140200711067520 basic_session_run_hooks.py:260] loss = 1.0764705, step = 184700 (3.262 sec)
I0804 21:13:28.235992 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.097
I0804 21:13:28.237291 140200711067520 basic_session_run_hooks.py:260] loss = 1.0917569, step = 184800 (3.215 sec)
I0804 21:13:31.460234 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0149
I0804 21:13:31.461683 140200711067520 basic_session_run_hooks.py:260] loss = 1.0752122, step = 184900 (3.224 sec)
I0804 21:13:34.627925 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 185000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:13:34.930472 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:13:34.972170 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4743
I0804 21:13:34.973803 140200711067520 basic_session_run_hooks.py:260] loss = 1.1508712, step = 185000 (3.512 sec)
I0804 21:13:38.172758 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2443
I0804 21:13:38.174300 140200711067520 basic_session_run_hooks.py:260] loss = 1.0084317, step = 185100 (3.200 sec)
I0804 21:13:41.389554 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.087
I0804 21:13:41.390874 140200711067520 basic_session_run_hooks.py:260] loss = 1.083597, step = 185200 (3.217 sec)
I0804 21:13:44.622186 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9345
I0804 21:13:44.623894 140200711067520 basic_session_run_hooks.py:260] loss = 1.1222187, step = 185300 (3.233 sec)
I0804 21:13:47.846663 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0127
I0804 21:13:47.847930 140200711067520 basic_session_run_hooks.py:260] loss = 1.05575, step = 185400 (3.224 sec)
I0804 21:13:51.058064 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1392
I0804 21:13:51.059147 140200711067520 basic_session_run_hooks.py:260] loss = 1.0435941, step = 185500 (3.211 sec)
I0804 21:13:54.268749 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1458
I0804 21:13:54.269856 140200711067520 basic_session_run_hooks.py:260] loss = 1.1384064, step = 185600 (3.211 sec)
I0804 21:13:57.500088 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.947
I0804 21:13:57.501495 140200711067520 basic_session_run_hooks.py:260] loss = 1.0165087, step = 185700 (3.232 sec)
I0804 21:14:00.683349 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4143
I0804 21:14:00.684795 140200711067520 basic_session_run_hooks.py:260] loss = 1.1734078, step = 185800 (3.183 sec)
I0804 21:14:03.879472 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2881
I0804 21:14:03.880706 140200711067520 basic_session_run_hooks.py:260] loss = 1.0499064, step = 185900 (3.196 sec)
I0804 21:14:07.038090 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 186000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:14:07.332257 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:14:07.370326 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6459
I0804 21:14:07.371395 140200711067520 basic_session_run_hooks.py:260] loss = 1.1116105, step = 186000 (3.491 sec)
I0804 21:14:10.562635 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3256
I0804 21:14:10.564017 140200711067520 basic_session_run_hooks.py:260] loss = 1.1165146, step = 186100 (3.193 sec)
I0804 21:14:13.755165 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3237
I0804 21:14:13.756447 140200711067520 basic_session_run_hooks.py:260] loss = 1.0247709, step = 186200 (3.192 sec)
I0804 21:14:16.927273 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5576
I0804 21:14:16.928891 140200711067520 basic_session_run_hooks.py:260] loss = 1.101238, step = 186300 (3.172 sec)
I0804 21:14:20.119355 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2946
I0804 21:14:20.120863 140200711067520 basic_session_run_hooks.py:260] loss = 1.1473997, step = 186400 (3.192 sec)
I0804 21:14:23.303383 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4067
I0804 21:14:23.304751 140200711067520 basic_session_run_hooks.py:260] loss = 1.0750734, step = 186500 (3.184 sec)
I0804 21:14:26.477945 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5006
I0804 21:14:26.479315 140200711067520 basic_session_run_hooks.py:260] loss = 1.1246649, step = 186600 (3.175 sec)
I0804 21:14:29.626644 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.759
I0804 21:14:29.628304 140200711067520 basic_session_run_hooks.py:260] loss = 1.1113755, step = 186700 (3.149 sec)
I0804 21:14:32.786974 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6424
I0804 21:14:32.788373 140200711067520 basic_session_run_hooks.py:260] loss = 1.0569437, step = 186800 (3.160 sec)
I0804 21:14:35.970636 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4102
I0804 21:14:35.971907 140200711067520 basic_session_run_hooks.py:260] loss = 1.0581504, step = 186900 (3.184 sec)
I0804 21:14:39.136250 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 187000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:14:39.430507 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:14:39.473814 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5455
I0804 21:14:39.475212 140200711067520 basic_session_run_hooks.py:260] loss = 1.1233642, step = 187000 (3.503 sec)
I0804 21:14:42.697127 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.024
I0804 21:14:42.698494 140200711067520 basic_session_run_hooks.py:260] loss = 1.1386685, step = 187100 (3.223 sec)
I0804 21:14:45.915336 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0732
I0804 21:14:45.916694 140200711067520 basic_session_run_hooks.py:260] loss = 1.1073948, step = 187200 (3.218 sec)
I0804 21:14:49.135804 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0537
I0804 21:14:49.137013 140200711067520 basic_session_run_hooks.py:260] loss = 1.0890274, step = 187300 (3.220 sec)
I0804 21:14:52.317800 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4245
I0804 21:14:52.319061 140200711067520 basic_session_run_hooks.py:260] loss = 1.0798773, step = 187400 (3.182 sec)
I0804 21:14:55.601033 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4577
I0804 21:14:55.602482 140200711067520 basic_session_run_hooks.py:260] loss = 1.0331433, step = 187500 (3.283 sec)
I0804 21:14:58.759627 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6598
I0804 21:14:58.760984 140200711067520 basic_session_run_hooks.py:260] loss = 1.0669309, step = 187600 (3.159 sec)
I0804 21:15:01.927936 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5626
I0804 21:15:01.929396 140200711067520 basic_session_run_hooks.py:260] loss = 1.063528, step = 187700 (3.168 sec)
I0804 21:15:05.118883 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3386
I0804 21:15:05.120310 140200711067520 basic_session_run_hooks.py:260] loss = 1.0727788, step = 187800 (3.191 sec)
I0804 21:15:08.303369 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4022
I0804 21:15:08.304874 140200711067520 basic_session_run_hooks.py:260] loss = 0.9905544, step = 187900 (3.185 sec)
I0804 21:15:11.456924 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 188000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:15:11.751856 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:15:11.796943 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6238
I0804 21:15:11.798172 140200711067520 basic_session_run_hooks.py:260] loss = 1.089804, step = 188000 (3.493 sec)
I0804 21:15:14.974430 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4717
I0804 21:15:14.975615 140200711067520 basic_session_run_hooks.py:260] loss = 1.0824281, step = 188100 (3.177 sec)
I0804 21:15:18.167050 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3223
I0804 21:15:18.168520 140200711067520 basic_session_run_hooks.py:260] loss = 1.1041087, step = 188200 (3.193 sec)
I0804 21:15:21.376274 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1601
I0804 21:15:21.377700 140200711067520 basic_session_run_hooks.py:260] loss = 1.058436, step = 188300 (3.209 sec)
I0804 21:15:24.597689 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0422
I0804 21:15:24.599064 140200711067520 basic_session_run_hooks.py:260] loss = 1.05537, step = 188400 (3.221 sec)
I0804 21:15:27.807550 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1543
I0804 21:15:27.808971 140200711067520 basic_session_run_hooks.py:260] loss = 1.1301793, step = 188500 (3.210 sec)
I0804 21:15:31.013799 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1889
I0804 21:15:31.015223 140200711067520 basic_session_run_hooks.py:260] loss = 1.0868847, step = 188600 (3.206 sec)
I0804 21:15:34.223438 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1562
I0804 21:15:34.224781 140200711067520 basic_session_run_hooks.py:260] loss = 1.0768968, step = 188700 (3.210 sec)
I0804 21:15:37.431654 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1698
I0804 21:15:37.432769 140200711067520 basic_session_run_hooks.py:260] loss = 1.0412124, step = 188800 (3.208 sec)
I0804 21:15:40.640138 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1673
I0804 21:15:40.641571 140200711067520 basic_session_run_hooks.py:260] loss = 1.1039554, step = 188900 (3.209 sec)
I0804 21:15:43.842915 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 189000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:15:44.150784 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:15:44.189905 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.1707
I0804 21:15:44.191107 140200711067520 basic_session_run_hooks.py:260] loss = 1.0046321, step = 189000 (3.550 sec)
I0804 21:15:47.332255 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8237
I0804 21:15:47.333743 140200711067520 basic_session_run_hooks.py:260] loss = 1.0946923, step = 189100 (3.143 sec)
I0804 21:15:50.459490 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9772
I0804 21:15:50.460745 140200711067520 basic_session_run_hooks.py:260] loss = 1.0821708, step = 189200 (3.127 sec)
I0804 21:15:53.584000 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 32.0049
I0804 21:15:53.585629 140200711067520 basic_session_run_hooks.py:260] loss = 1.1885908, step = 189300 (3.125 sec)
I0804 21:15:56.747675 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.609
I0804 21:15:56.749050 140200711067520 basic_session_run_hooks.py:260] loss = 1.0570059, step = 189400 (3.163 sec)
I0804 21:15:59.887733 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8463
I0804 21:15:59.888971 140200711067520 basic_session_run_hooks.py:260] loss = 1.1570084, step = 189500 (3.140 sec)
I0804 21:16:03.024357 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8815
I0804 21:16:03.025876 140200711067520 basic_session_run_hooks.py:260] loss = 1.1196383, step = 189600 (3.137 sec)
I0804 21:16:06.156674 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9251
I0804 21:16:06.158160 140200711067520 basic_session_run_hooks.py:260] loss = 1.019074, step = 189700 (3.132 sec)
I0804 21:16:09.325328 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5592
I0804 21:16:09.326864 140200711067520 basic_session_run_hooks.py:260] loss = 1.1260017, step = 189800 (3.169 sec)
I0804 21:16:12.514038 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3607
I0804 21:16:12.515506 140200711067520 basic_session_run_hooks.py:260] loss = 1.1006281, step = 189900 (3.189 sec)
I0804 21:16:15.662985 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 190000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:16:15.973246 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:16:16.017677 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5414
I0804 21:16:16.018649 140200711067520 basic_session_run_hooks.py:260] loss = 0.99487877, step = 190000 (3.503 sec)
I0804 21:16:19.202374 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4006
I0804 21:16:19.203580 140200711067520 basic_session_run_hooks.py:260] loss = 1.1535513, step = 190100 (3.185 sec)
I0804 21:16:22.383172 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4385
I0804 21:16:22.384265 140200711067520 basic_session_run_hooks.py:260] loss = 1.0546324, step = 190200 (3.181 sec)
I0804 21:16:25.564482 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.434
I0804 21:16:25.565721 140200711067520 basic_session_run_hooks.py:260] loss = 1.0771601, step = 190300 (3.181 sec)
I0804 21:16:28.756547 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3277
I0804 21:16:28.758584 140200711067520 basic_session_run_hooks.py:260] loss = 1.056657, step = 190400 (3.193 sec)
I0804 21:16:31.960197 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2143
I0804 21:16:31.961534 140200711067520 basic_session_run_hooks.py:260] loss = 1.022684, step = 190500 (3.203 sec)
I0804 21:16:35.214993 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7238
I0804 21:16:35.216130 140200711067520 basic_session_run_hooks.py:260] loss = 1.0662925, step = 190600 (3.255 sec)
I0804 21:16:38.440533 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0027
I0804 21:16:38.441689 140200711067520 basic_session_run_hooks.py:260] loss = 1.166308, step = 190700 (3.226 sec)
I0804 21:16:41.660345 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0577
I0804 21:16:41.661941 140200711067520 basic_session_run_hooks.py:260] loss = 0.97773093, step = 190800 (3.220 sec)
I0804 21:16:44.852495 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3268
I0804 21:16:44.853849 140200711067520 basic_session_run_hooks.py:260] loss = 1.0939617, step = 190900 (3.192 sec)
I0804 21:16:47.995119 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 191000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:16:48.308164 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 21:16:48.309448 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 21:16:48.455092 140200711067520 estimator.py:1145] Calling model_fn.
I0804 21:16:48.456027 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 21:16:48.456442 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 21:16:48.456539 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 21:16:48.456622 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 21:16:48.456696 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 21:16:48.456780 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 21:16:48.456846 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 21:16:48.549839 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 21:16:48.609145 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 21:16:48.749556 140200711067520 t2t_model.py:2172] Building model body
I0804 21:16:49.700961 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 21:16:50.399734 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 21:16:50.417544 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T21:16:50Z
I0804 21:16:50.581239 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 21:16:50.581892: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:16:50.582287: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 21:16:50.582390: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 21:16:50.582415: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 21:16:50.582453: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 21:16:50.582475: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 21:16:50.582496: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 21:16:50.582518: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 21:16:50.582541: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 21:16:50.582642: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:16:50.583039: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:16:50.583356: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 21:16:50.583399: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 21:16:50.583412: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 21:16:50.583438: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 21:16:50.583717: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:16:50.584129: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:16:50.584474: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 21:16:50.586051 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-191000
I0804 21:16:50.801615 140200711067520 session_manager.py:500] Running local_init_op.
I0804 21:16:50.846954 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 21:16:56.883124 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 21:17:02.185722 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 21:17:07.490707 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 21:17:12.782867 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 21:17:18.057094 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 21:17:23.383639 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 21:17:28.686561 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 21:17:34.024442 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 21:17:39.316114 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 21:17:44.398901 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-21:17:44
I0804 21:17:44.399133 140200711067520 estimator.py:2039] Saving dict for global step 191000: global_step = 191000, loss = 1.1653775, metrics-paper_generation_problem/targets/accuracy = 0.6766576, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.886202, metrics-paper_generation_problem/targets/approx_bleu_score = 0.4940288, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1654139, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5881068, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.698595
I0804 21:17:44.399597 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 191000: experiment/transformer/transformer_small/output/model.ckpt-191000
I0804 21:17:44.458705 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.67768
I0804 21:17:44.459879 140200711067520 basic_session_run_hooks.py:260] loss = 1.0061302, step = 191000 (59.606 sec)
I0804 21:17:47.705229 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8026
I0804 21:17:47.706544 140200711067520 basic_session_run_hooks.py:260] loss = 1.0946387, step = 191100 (3.247 sec)
I0804 21:17:50.896675 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3337
I0804 21:17:50.897999 140200711067520 basic_session_run_hooks.py:260] loss = 0.9863627, step = 191200 (3.191 sec)
I0804 21:17:54.108469 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1353
I0804 21:17:54.109906 140200711067520 basic_session_run_hooks.py:260] loss = 1.0924733, step = 191300 (3.212 sec)
I0804 21:17:57.348777 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8612
I0804 21:17:57.350258 140200711067520 basic_session_run_hooks.py:260] loss = 1.0605073, step = 191400 (3.240 sec)
I0804 21:18:00.521076 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5229
I0804 21:18:00.522787 140200711067520 basic_session_run_hooks.py:260] loss = 1.1032586, step = 191500 (3.173 sec)
I0804 21:18:03.696153 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4953
I0804 21:18:03.697439 140200711067520 basic_session_run_hooks.py:260] loss = 1.1253883, step = 191600 (3.175 sec)
I0804 21:18:06.897643 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2354
I0804 21:18:06.899178 140200711067520 basic_session_run_hooks.py:260] loss = 1.179381, step = 191700 (3.202 sec)
I0804 21:18:10.118636 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0464
I0804 21:18:10.119876 140200711067520 basic_session_run_hooks.py:260] loss = 1.0831317, step = 191800 (3.221 sec)
I0804 21:18:13.328831 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1508
I0804 21:18:13.330186 140200711067520 basic_session_run_hooks.py:260] loss = 1.1220344, step = 191900 (3.210 sec)
I0804 21:18:16.475574 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 192000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:18:16.783117 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:18:16.821192 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6336
I0804 21:18:16.822388 140200711067520 basic_session_run_hooks.py:260] loss = 1.0971903, step = 192000 (3.492 sec)
I0804 21:18:19.995674 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5015
I0804 21:18:19.997000 140200711067520 basic_session_run_hooks.py:260] loss = 1.1190305, step = 192100 (3.175 sec)
I0804 21:18:23.195312 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2536
I0804 21:18:23.196646 140200711067520 basic_session_run_hooks.py:260] loss = 1.0394591, step = 192200 (3.200 sec)
I0804 21:18:26.376502 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4349
I0804 21:18:26.377956 140200711067520 basic_session_run_hooks.py:260] loss = 0.9965354, step = 192300 (3.181 sec)
I0804 21:18:29.531832 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6921
I0804 21:18:29.533108 140200711067520 basic_session_run_hooks.py:260] loss = 1.0451137, step = 192400 (3.155 sec)
I0804 21:18:32.689327 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6707
I0804 21:18:32.690832 140200711067520 basic_session_run_hooks.py:260] loss = 1.0419977, step = 192500 (3.158 sec)
I0804 21:18:35.836512 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7748
I0804 21:18:35.837880 140200711067520 basic_session_run_hooks.py:260] loss = 1.0920428, step = 192600 (3.147 sec)
I0804 21:18:38.978993 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8217
I0804 21:18:38.980553 140200711067520 basic_session_run_hooks.py:260] loss = 1.0496783, step = 192700 (3.143 sec)
I0804 21:18:42.119318 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8438
I0804 21:18:42.120654 140200711067520 basic_session_run_hooks.py:260] loss = 1.0937288, step = 192800 (3.140 sec)
I0804 21:18:45.267351 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7659
I0804 21:18:45.269068 140200711067520 basic_session_run_hooks.py:260] loss = 1.1558946, step = 192900 (3.148 sec)
I0804 21:18:48.451083 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 193000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:18:48.746053 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:18:48.789646 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3903
I0804 21:18:48.790714 140200711067520 basic_session_run_hooks.py:260] loss = 0.98306364, step = 193000 (3.522 sec)
I0804 21:18:51.976712 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3772
I0804 21:18:51.977964 140200711067520 basic_session_run_hooks.py:260] loss = 1.1312164, step = 193100 (3.187 sec)
I0804 21:18:55.179844 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2194
I0804 21:18:55.181334 140200711067520 basic_session_run_hooks.py:260] loss = 1.0734868, step = 193200 (3.203 sec)
I0804 21:18:58.379507 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2533
I0804 21:18:58.380787 140200711067520 basic_session_run_hooks.py:260] loss = 1.1507102, step = 193300 (3.199 sec)
I0804 21:19:01.578628 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2584
I0804 21:19:01.579947 140200711067520 basic_session_run_hooks.py:260] loss = 1.019576, step = 193400 (3.199 sec)
I0804 21:19:04.810661 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9404
I0804 21:19:04.811810 140200711067520 basic_session_run_hooks.py:260] loss = 1.0855082, step = 193500 (3.232 sec)
I0804 21:19:08.064309 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7348
I0804 21:19:08.065398 140200711067520 basic_session_run_hooks.py:260] loss = 1.086294, step = 193600 (3.254 sec)
I0804 21:19:11.283705 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0617
I0804 21:19:11.284897 140200711067520 basic_session_run_hooks.py:260] loss = 1.1395323, step = 193700 (3.219 sec)
I0804 21:19:14.481689 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2698
I0804 21:19:14.482822 140200711067520 basic_session_run_hooks.py:260] loss = 1.1056229, step = 193800 (3.198 sec)
I0804 21:19:17.617754 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.887
I0804 21:19:17.618867 140200711067520 basic_session_run_hooks.py:260] loss = 1.0633337, step = 193900 (3.136 sec)
I0804 21:19:20.743579 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 194000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:19:21.040922 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:19:21.080113 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8818
I0804 21:19:21.081358 140200711067520 basic_session_run_hooks.py:260] loss = 1.0259215, step = 194000 (3.462 sec)
I0804 21:19:24.244409 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6029
I0804 21:19:24.245624 140200711067520 basic_session_run_hooks.py:260] loss = 0.99994206, step = 194100 (3.164 sec)
I0804 21:19:27.415196 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.538
I0804 21:19:27.416471 140200711067520 basic_session_run_hooks.py:260] loss = 1.0624233, step = 194200 (3.171 sec)
I0804 21:19:30.585782 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5399
I0804 21:19:30.587103 140200711067520 basic_session_run_hooks.py:260] loss = 1.1015216, step = 194300 (3.171 sec)
I0804 21:19:33.739380 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7097
I0804 21:19:33.741017 140200711067520 basic_session_run_hooks.py:260] loss = 1.0709002, step = 194400 (3.154 sec)
I0804 21:19:36.912782 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.512
I0804 21:19:36.914002 140200711067520 basic_session_run_hooks.py:260] loss = 1.0788685, step = 194500 (3.173 sec)
I0804 21:19:40.134915 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0353
I0804 21:19:40.136049 140200711067520 basic_session_run_hooks.py:260] loss = 0.9873385, step = 194600 (3.222 sec)
I0804 21:19:43.356932 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0364
I0804 21:19:43.358134 140200711067520 basic_session_run_hooks.py:260] loss = 1.0697428, step = 194700 (3.222 sec)
I0804 21:19:46.547379 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3435
I0804 21:19:46.548755 140200711067520 basic_session_run_hooks.py:260] loss = 1.0710206, step = 194800 (3.191 sec)
I0804 21:19:49.764620 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0826
I0804 21:19:49.765939 140200711067520 basic_session_run_hooks.py:260] loss = 1.0421448, step = 194900 (3.217 sec)
I0804 21:19:52.914026 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 195000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:19:53.214686 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:19:53.254746 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6521
I0804 21:19:53.255871 140200711067520 basic_session_run_hooks.py:260] loss = 1.0710396, step = 195000 (3.490 sec)
I0804 21:19:56.447983 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3166
I0804 21:19:56.449448 140200711067520 basic_session_run_hooks.py:260] loss = 1.0659773, step = 195100 (3.194 sec)
I0804 21:19:59.635256 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3747
I0804 21:19:59.636475 140200711067520 basic_session_run_hooks.py:260] loss = 1.0160112, step = 195200 (3.187 sec)
I0804 21:20:02.847466 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1314
I0804 21:20:02.848892 140200711067520 basic_session_run_hooks.py:260] loss = 1.0871261, step = 195300 (3.212 sec)
I0804 21:20:06.043154 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.292
I0804 21:20:06.044546 140200711067520 basic_session_run_hooks.py:260] loss = 1.0416194, step = 195400 (3.196 sec)
I0804 21:20:09.216664 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5108
I0804 21:20:09.217993 140200711067520 basic_session_run_hooks.py:260] loss = 1.045076, step = 195500 (3.173 sec)
I0804 21:20:12.387705 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5355
I0804 21:20:12.389119 140200711067520 basic_session_run_hooks.py:260] loss = 0.9817378, step = 195600 (3.171 sec)
I0804 21:20:15.555142 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5711
I0804 21:20:15.556641 140200711067520 basic_session_run_hooks.py:260] loss = 0.9993578, step = 195700 (3.168 sec)
I0804 21:20:18.739250 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4061
I0804 21:20:18.740876 140200711067520 basic_session_run_hooks.py:260] loss = 1.0160722, step = 195800 (3.184 sec)
I0804 21:20:21.930238 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3382
I0804 21:20:21.931710 140200711067520 basic_session_run_hooks.py:260] loss = 1.0812558, step = 195900 (3.191 sec)
I0804 21:20:25.069093 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 196000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:20:25.370240 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:20:25.409121 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7446
I0804 21:20:25.410048 140200711067520 basic_session_run_hooks.py:260] loss = 1.0268708, step = 196000 (3.478 sec)
I0804 21:20:28.597390 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3653
I0804 21:20:28.598987 140200711067520 basic_session_run_hooks.py:260] loss = 1.1323452, step = 196100 (3.189 sec)
I0804 21:20:31.765370 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5659
I0804 21:20:31.766742 140200711067520 basic_session_run_hooks.py:260] loss = 1.0964614, step = 196200 (3.168 sec)
I0804 21:20:34.972779 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.178
I0804 21:20:34.973936 140200711067520 basic_session_run_hooks.py:260] loss = 1.1114868, step = 196300 (3.207 sec)
I0804 21:20:38.142588 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5477
I0804 21:20:38.143736 140200711067520 basic_session_run_hooks.py:260] loss = 1.0065557, step = 196400 (3.170 sec)
I0804 21:20:41.307990 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5915
I0804 21:20:41.309338 140200711067520 basic_session_run_hooks.py:260] loss = 1.1310524, step = 196500 (3.166 sec)
I0804 21:20:44.483894 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4871
I0804 21:20:44.485134 140200711067520 basic_session_run_hooks.py:260] loss = 1.0656706, step = 196600 (3.176 sec)
I0804 21:20:47.659098 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4939
I0804 21:20:47.660597 140200711067520 basic_session_run_hooks.py:260] loss = 1.0345689, step = 196700 (3.175 sec)
I0804 21:20:50.837509 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4626
I0804 21:20:50.838839 140200711067520 basic_session_run_hooks.py:260] loss = 1.1271529, step = 196800 (3.178 sec)
I0804 21:20:54.026836 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3543
I0804 21:20:54.027957 140200711067520 basic_session_run_hooks.py:260] loss = 1.0608749, step = 196900 (3.189 sec)
I0804 21:20:57.171811 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 197000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:20:57.471443 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:20:57.517226 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6499
I0804 21:20:57.518259 140200711067520 basic_session_run_hooks.py:260] loss = 1.0474122, step = 197000 (3.490 sec)
I0804 21:21:00.692214 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4966
I0804 21:21:00.693764 140200711067520 basic_session_run_hooks.py:260] loss = 1.0970184, step = 197100 (3.176 sec)
I0804 21:21:03.870261 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4658
I0804 21:21:03.872035 140200711067520 basic_session_run_hooks.py:260] loss = 1.0099502, step = 197200 (3.178 sec)
I0804 21:21:07.183148 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.1853
I0804 21:21:07.184465 140200711067520 basic_session_run_hooks.py:260] loss = 1.0442789, step = 197300 (3.312 sec)
I0804 21:21:10.377437 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3059
I0804 21:21:10.378925 140200711067520 basic_session_run_hooks.py:260] loss = 1.0953308, step = 197400 (3.194 sec)
I0804 21:21:13.594025 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0887
I0804 21:21:13.595338 140200711067520 basic_session_run_hooks.py:260] loss = 1.042183, step = 197500 (3.216 sec)
I0804 21:21:16.775549 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4317
I0804 21:21:16.776736 140200711067520 basic_session_run_hooks.py:260] loss = 1.0211613, step = 197600 (3.181 sec)
I0804 21:21:19.954157 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4601
I0804 21:21:19.955675 140200711067520 basic_session_run_hooks.py:260] loss = 1.1160389, step = 197700 (3.179 sec)
I0804 21:21:23.170441 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0919
I0804 21:21:23.171736 140200711067520 basic_session_run_hooks.py:260] loss = 1.0592711, step = 197800 (3.216 sec)
I0804 21:21:26.381250 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1447
I0804 21:21:26.382342 140200711067520 basic_session_run_hooks.py:260] loss = 1.067345, step = 197900 (3.211 sec)
I0804 21:21:29.556779 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 198000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:21:29.850392 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:21:29.890750 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4939
I0804 21:21:29.891770 140200711067520 basic_session_run_hooks.py:260] loss = 1.0280184, step = 198000 (3.509 sec)
I0804 21:21:33.115992 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0057
I0804 21:21:33.117087 140200711067520 basic_session_run_hooks.py:260] loss = 1.1249108, step = 198100 (3.225 sec)
I0804 21:21:36.301499 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3925
I0804 21:21:36.302947 140200711067520 basic_session_run_hooks.py:260] loss = 1.0090152, step = 198200 (3.186 sec)
I0804 21:21:39.512798 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1396
I0804 21:21:39.514156 140200711067520 basic_session_run_hooks.py:260] loss = 1.0500352, step = 198300 (3.211 sec)
I0804 21:21:42.726007 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1216
I0804 21:21:42.727395 140200711067520 basic_session_run_hooks.py:260] loss = 1.1282649, step = 198400 (3.213 sec)
I0804 21:21:45.956582 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9544
I0804 21:21:45.958159 140200711067520 basic_session_run_hooks.py:260] loss = 1.0500304, step = 198500 (3.231 sec)
I0804 21:21:49.197300 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8572
I0804 21:21:49.198880 140200711067520 basic_session_run_hooks.py:260] loss = 1.0885893, step = 198600 (3.241 sec)
I0804 21:21:52.433022 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9049
I0804 21:21:52.434347 140200711067520 basic_session_run_hooks.py:260] loss = 1.0162288, step = 198700 (3.235 sec)
I0804 21:21:55.632148 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2586
I0804 21:21:55.633520 140200711067520 basic_session_run_hooks.py:260] loss = 1.0677074, step = 198800 (3.199 sec)
I0804 21:21:58.858551 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9944
I0804 21:21:58.859961 140200711067520 basic_session_run_hooks.py:260] loss = 1.0852252, step = 198900 (3.226 sec)
I0804 21:22:02.047649 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 199000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:22:02.347621 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:22:02.390380 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3136
I0804 21:22:02.391463 140200711067520 basic_session_run_hooks.py:260] loss = 1.084918, step = 199000 (3.531 sec)
I0804 21:22:05.597167 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1841
I0804 21:22:05.598594 140200711067520 basic_session_run_hooks.py:260] loss = 1.0324137, step = 199100 (3.207 sec)
I0804 21:22:08.801440 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2106
I0804 21:22:08.802782 140200711067520 basic_session_run_hooks.py:260] loss = 1.0906309, step = 199200 (3.204 sec)
I0804 21:22:12.006527 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1983
I0804 21:22:12.007762 140200711067520 basic_session_run_hooks.py:260] loss = 1.0369992, step = 199300 (3.205 sec)
I0804 21:22:15.225287 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0677
I0804 21:22:15.226916 140200711067520 basic_session_run_hooks.py:260] loss = 1.0169804, step = 199400 (3.219 sec)
I0804 21:22:18.415159 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3492
I0804 21:22:18.416836 140200711067520 basic_session_run_hooks.py:260] loss = 1.0959753, step = 199500 (3.190 sec)
I0804 21:22:21.605286 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3467
I0804 21:22:21.606605 140200711067520 basic_session_run_hooks.py:260] loss = 1.0766824, step = 199600 (3.190 sec)
I0804 21:22:24.827524 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0346
I0804 21:22:24.828721 140200711067520 basic_session_run_hooks.py:260] loss = 1.134889, step = 199700 (3.222 sec)
I0804 21:22:28.010643 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4157
I0804 21:22:28.012015 140200711067520 basic_session_run_hooks.py:260] loss = 1.019222, step = 199800 (3.183 sec)
I0804 21:22:31.159734 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7552
I0804 21:22:31.160922 140200711067520 basic_session_run_hooks.py:260] loss = 1.0882303, step = 199900 (3.149 sec)
I0804 21:22:34.292789 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 200000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:22:34.590018 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:22:34.629517 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8201
I0804 21:22:34.630621 140200711067520 basic_session_run_hooks.py:260] loss = 1.0636154, step = 200000 (3.470 sec)
I0804 21:22:37.820285 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3405
I0804 21:22:37.821702 140200711067520 basic_session_run_hooks.py:260] loss = 1.0253745, step = 200100 (3.191 sec)
I0804 21:22:40.968592 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7632
I0804 21:22:40.969901 140200711067520 basic_session_run_hooks.py:260] loss = 1.0438075, step = 200200 (3.148 sec)
I0804 21:22:44.121081 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7209
I0804 21:22:44.122397 140200711067520 basic_session_run_hooks.py:260] loss = 1.0507977, step = 200300 (3.152 sec)
I0804 21:22:47.272781 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7289
I0804 21:22:47.274162 140200711067520 basic_session_run_hooks.py:260] loss = 1.0359304, step = 200400 (3.152 sec)
I0804 21:22:50.480207 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1777
I0804 21:22:50.481547 140200711067520 basic_session_run_hooks.py:260] loss = 0.9729346, step = 200500 (3.207 sec)
I0804 21:22:53.669835 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3514
I0804 21:22:53.671010 140200711067520 basic_session_run_hooks.py:260] loss = 1.1163907, step = 200600 (3.189 sec)
I0804 21:22:56.875792 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.192
I0804 21:22:56.876954 140200711067520 basic_session_run_hooks.py:260] loss = 0.9992767, step = 200700 (3.206 sec)
I0804 21:23:00.073136 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2759
I0804 21:23:00.074311 140200711067520 basic_session_run_hooks.py:260] loss = 1.055311, step = 200800 (3.197 sec)
I0804 21:23:03.296220 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0263
I0804 21:23:03.297395 140200711067520 basic_session_run_hooks.py:260] loss = 0.9907532, step = 200900 (3.223 sec)
I0804 21:23:06.464148 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 201000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:23:06.775672 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:23:06.821404 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3671
I0804 21:23:06.822684 140200711067520 basic_session_run_hooks.py:260] loss = 1.0688014, step = 201000 (3.525 sec)
I0804 21:23:10.017749 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.286
I0804 21:23:10.019076 140200711067520 basic_session_run_hooks.py:260] loss = 1.0438105, step = 201100 (3.196 sec)
I0804 21:23:13.223575 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1934
I0804 21:23:13.224846 140200711067520 basic_session_run_hooks.py:260] loss = 1.0250725, step = 201200 (3.206 sec)
I0804 21:23:16.451757 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9769
I0804 21:23:16.452891 140200711067520 basic_session_run_hooks.py:260] loss = 1.0336745, step = 201300 (3.228 sec)
I0804 21:23:19.648357 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2834
I0804 21:23:19.649832 140200711067520 basic_session_run_hooks.py:260] loss = 1.0371219, step = 201400 (3.197 sec)
I0804 21:23:22.820548 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5241
I0804 21:23:22.821692 140200711067520 basic_session_run_hooks.py:260] loss = 0.995201, step = 201500 (3.172 sec)
I0804 21:23:25.996903 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4823
I0804 21:23:25.998488 140200711067520 basic_session_run_hooks.py:260] loss = 1.1711582, step = 201600 (3.177 sec)
I0804 21:23:29.167752 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5374
I0804 21:23:29.169016 140200711067520 basic_session_run_hooks.py:260] loss = 1.0354856, step = 201700 (3.171 sec)
I0804 21:23:32.336342 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5597
I0804 21:23:32.337708 140200711067520 basic_session_run_hooks.py:260] loss = 1.0607649, step = 201800 (3.169 sec)
I0804 21:23:35.511215 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4973
I0804 21:23:35.512820 140200711067520 basic_session_run_hooks.py:260] loss = 1.0602365, step = 201900 (3.175 sec)
I0804 21:23:38.629928 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 202000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:23:38.923026 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:23:38.959470 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.0002
I0804 21:23:38.960653 140200711067520 basic_session_run_hooks.py:260] loss = 1.053769, step = 202000 (3.448 sec)
I0804 21:23:42.147221 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3703
I0804 21:23:42.148606 140200711067520 basic_session_run_hooks.py:260] loss = 1.0557011, step = 202100 (3.188 sec)
I0804 21:23:45.298825 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7298
I0804 21:23:45.300034 140200711067520 basic_session_run_hooks.py:260] loss = 1.0457883, step = 202200 (3.151 sec)
I0804 21:23:48.444674 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7879
I0804 21:23:48.445976 140200711067520 basic_session_run_hooks.py:260] loss = 1.1160995, step = 202300 (3.146 sec)
I0804 21:23:51.600956 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.683
I0804 21:23:51.602335 140200711067520 basic_session_run_hooks.py:260] loss = 1.0539001, step = 202400 (3.156 sec)
I0804 21:23:54.727436 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.985
I0804 21:23:54.728939 140200711067520 basic_session_run_hooks.py:260] loss = 1.083011, step = 202500 (3.127 sec)
I0804 21:23:57.885056 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6692
I0804 21:23:57.886413 140200711067520 basic_session_run_hooks.py:260] loss = 1.0581592, step = 202600 (3.157 sec)
I0804 21:24:01.057727 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5192
I0804 21:24:01.059287 140200711067520 basic_session_run_hooks.py:260] loss = 1.0752395, step = 202700 (3.173 sec)
I0804 21:24:04.201589 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8082
I0804 21:24:04.202829 140200711067520 basic_session_run_hooks.py:260] loss = 1.0008317, step = 202800 (3.144 sec)
I0804 21:24:07.406071 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2061
I0804 21:24:07.407407 140200711067520 basic_session_run_hooks.py:260] loss = 1.0801786, step = 202900 (3.205 sec)
I0804 21:24:10.592720 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 203000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:24:10.887673 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:24:10.926568 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.405
I0804 21:24:10.927535 140200711067520 basic_session_run_hooks.py:260] loss = 1.0820597, step = 203000 (3.520 sec)
I0804 21:24:14.142041 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0997
I0804 21:24:14.143497 140200711067520 basic_session_run_hooks.py:260] loss = 0.9845524, step = 203100 (3.216 sec)
I0804 21:24:17.329581 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3724
I0804 21:24:17.331076 140200711067520 basic_session_run_hooks.py:260] loss = 1.1231854, step = 203200 (3.188 sec)
I0804 21:24:20.519788 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3456
I0804 21:24:20.520855 140200711067520 basic_session_run_hooks.py:260] loss = 1.1265386, step = 203300 (3.190 sec)
I0804 21:24:23.720857 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2397
I0804 21:24:23.722330 140200711067520 basic_session_run_hooks.py:260] loss = 1.0924197, step = 203400 (3.201 sec)
I0804 21:24:26.937754 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0858
I0804 21:24:26.938915 140200711067520 basic_session_run_hooks.py:260] loss = 1.0662233, step = 203500 (3.217 sec)
I0804 21:24:30.158519 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0487
I0804 21:24:30.159738 140200711067520 basic_session_run_hooks.py:260] loss = 1.0181272, step = 203600 (3.221 sec)
I0804 21:24:33.407027 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7833
I0804 21:24:33.408471 140200711067520 basic_session_run_hooks.py:260] loss = 0.95900255, step = 203700 (3.249 sec)
I0804 21:24:36.618514 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1383
I0804 21:24:36.619874 140200711067520 basic_session_run_hooks.py:260] loss = 1.0486104, step = 203800 (3.211 sec)
I0804 21:24:39.812569 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3081
I0804 21:24:39.814063 140200711067520 basic_session_run_hooks.py:260] loss = 1.0465269, step = 203900 (3.194 sec)
I0804 21:24:42.996626 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 204000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:24:43.290640 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:24:43.330120 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4285
I0804 21:24:43.331054 140200711067520 basic_session_run_hooks.py:260] loss = 1.0550798, step = 204000 (3.517 sec)
I0804 21:24:46.569488 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8708
I0804 21:24:46.570761 140200711067520 basic_session_run_hooks.py:260] loss = 1.0041236, step = 204100 (3.240 sec)
I0804 21:24:49.795939 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9937
I0804 21:24:49.797000 140200711067520 basic_session_run_hooks.py:260] loss = 1.0462617, step = 204200 (3.226 sec)
I0804 21:24:52.987122 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3362
I0804 21:24:52.988235 140200711067520 basic_session_run_hooks.py:260] loss = 1.0669595, step = 204300 (3.191 sec)
I0804 21:24:56.207712 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0503
I0804 21:24:56.209010 140200711067520 basic_session_run_hooks.py:260] loss = 1.0643322, step = 204400 (3.221 sec)
I0804 21:24:59.398637 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3388
I0804 21:24:59.399922 140200711067520 basic_session_run_hooks.py:260] loss = 1.0751393, step = 204500 (3.191 sec)
I0804 21:25:02.539326 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8401
I0804 21:25:02.540677 140200711067520 basic_session_run_hooks.py:260] loss = 1.0881233, step = 204600 (3.141 sec)
I0804 21:25:05.679312 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8471
I0804 21:25:05.680434 140200711067520 basic_session_run_hooks.py:260] loss = 1.0532576, step = 204700 (3.140 sec)
I0804 21:25:08.827661 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7629
I0804 21:25:08.829105 140200711067520 basic_session_run_hooks.py:260] loss = 1.1140361, step = 204800 (3.149 sec)
I0804 21:25:11.977382 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7488
I0804 21:25:11.978839 140200711067520 basic_session_run_hooks.py:260] loss = 1.0294828, step = 204900 (3.150 sec)
I0804 21:25:15.098537 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 205000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:25:15.393380 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:25:15.434870 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.9225
I0804 21:25:15.435983 140200711067520 basic_session_run_hooks.py:260] loss = 0.9942222, step = 205000 (3.457 sec)
I0804 21:25:18.598503 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6097
I0804 21:25:18.599857 140200711067520 basic_session_run_hooks.py:260] loss = 1.0653806, step = 205100 (3.164 sec)
I0804 21:25:21.747216 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7589
I0804 21:25:21.748541 140200711067520 basic_session_run_hooks.py:260] loss = 1.0681946, step = 205200 (3.149 sec)
I0804 21:25:24.970783 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0215
I0804 21:25:24.971927 140200711067520 basic_session_run_hooks.py:260] loss = 1.0039834, step = 205300 (3.223 sec)
I0804 21:25:28.176248 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1967
I0804 21:25:28.177674 140200711067520 basic_session_run_hooks.py:260] loss = 1.0416933, step = 205400 (3.206 sec)
I0804 21:25:31.385275 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1621
I0804 21:25:31.386469 140200711067520 basic_session_run_hooks.py:260] loss = 1.0419832, step = 205500 (3.209 sec)
I0804 21:25:34.587333 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2298
I0804 21:25:34.588754 140200711067520 basic_session_run_hooks.py:260] loss = 1.130958, step = 205600 (3.202 sec)
I0804 21:25:37.795458 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.171
I0804 21:25:37.796738 140200711067520 basic_session_run_hooks.py:260] loss = 1.0311191, step = 205700 (3.208 sec)
I0804 21:25:40.996467 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2402
I0804 21:25:40.998107 140200711067520 basic_session_run_hooks.py:260] loss = 1.0537728, step = 205800 (3.201 sec)
I0804 21:25:44.218748 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0337
I0804 21:25:44.219974 140200711067520 basic_session_run_hooks.py:260] loss = 1.0694484, step = 205900 (3.222 sec)
I0804 21:25:47.415617 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 206000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:25:47.714228 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:25:47.752652 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2973
I0804 21:25:47.753905 140200711067520 basic_session_run_hooks.py:260] loss = 1.0013107, step = 206000 (3.534 sec)
I0804 21:25:50.971359 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0684
I0804 21:25:50.972650 140200711067520 basic_session_run_hooks.py:260] loss = 1.1938875, step = 206100 (3.219 sec)
I0804 21:25:54.171245 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2512
I0804 21:25:54.172353 140200711067520 basic_session_run_hooks.py:260] loss = 1.0837841, step = 206200 (3.200 sec)
I0804 21:25:57.382212 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1435
I0804 21:25:57.383531 140200711067520 basic_session_run_hooks.py:260] loss = 1.0600058, step = 206300 (3.211 sec)
I0804 21:26:00.576279 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3078
I0804 21:26:00.577721 140200711067520 basic_session_run_hooks.py:260] loss = 1.0186465, step = 206400 (3.194 sec)
I0804 21:26:03.807311 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9497
I0804 21:26:03.808743 140200711067520 basic_session_run_hooks.py:260] loss = 1.0520475, step = 206500 (3.231 sec)
I0804 21:26:07.030935 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0211
I0804 21:26:07.032016 140200711067520 basic_session_run_hooks.py:260] loss = 1.1143177, step = 206600 (3.223 sec)
I0804 21:26:10.243701 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1259
I0804 21:26:10.244927 140200711067520 basic_session_run_hooks.py:260] loss = 1.1734234, step = 206700 (3.213 sec)
I0804 21:26:13.459612 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0952
I0804 21:26:13.460715 140200711067520 basic_session_run_hooks.py:260] loss = 1.0739877, step = 206800 (3.216 sec)
I0804 21:26:16.620734 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6344
I0804 21:26:16.621973 140200711067520 basic_session_run_hooks.py:260] loss = 1.0245624, step = 206900 (3.161 sec)
I0804 21:26:19.754719 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 207000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:26:20.058157 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:26:20.095051 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7823
I0804 21:26:20.096107 140200711067520 basic_session_run_hooks.py:260] loss = 1.1209346, step = 207000 (3.474 sec)
I0804 21:26:23.243804 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.759
I0804 21:26:23.245200 140200711067520 basic_session_run_hooks.py:260] loss = 1.0788656, step = 207100 (3.149 sec)
I0804 21:26:26.482191 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8795
I0804 21:26:26.483569 140200711067520 basic_session_run_hooks.py:260] loss = 1.1092812, step = 207200 (3.238 sec)
I0804 21:26:29.671061 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3591
I0804 21:26:29.672630 140200711067520 basic_session_run_hooks.py:260] loss = 1.0750504, step = 207300 (3.189 sec)
I0804 21:26:32.853862 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4187
I0804 21:26:32.855321 140200711067520 basic_session_run_hooks.py:260] loss = 1.0730976, step = 207400 (3.183 sec)
I0804 21:26:36.053922 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2495
I0804 21:26:36.055608 140200711067520 basic_session_run_hooks.py:260] loss = 1.050307, step = 207500 (3.200 sec)
I0804 21:26:39.256706 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2229
I0804 21:26:39.258090 140200711067520 basic_session_run_hooks.py:260] loss = 1.0789127, step = 207600 (3.202 sec)
I0804 21:26:42.461207 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.206
I0804 21:26:42.462278 140200711067520 basic_session_run_hooks.py:260] loss = 1.025226, step = 207700 (3.204 sec)
I0804 21:26:45.654608 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3146
I0804 21:26:45.655937 140200711067520 basic_session_run_hooks.py:260] loss = 1.0388247, step = 207800 (3.194 sec)
I0804 21:26:48.881566 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9892
I0804 21:26:48.882941 140200711067520 basic_session_run_hooks.py:260] loss = 1.0659609, step = 207900 (3.227 sec)
I0804 21:26:52.075457 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 208000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:26:52.383342 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 21:26:52.384669 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 21:26:52.530561 140200711067520 estimator.py:1145] Calling model_fn.
I0804 21:26:52.531574 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 21:26:52.531960 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 21:26:52.532055 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 21:26:52.532136 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 21:26:52.532202 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 21:26:52.532285 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 21:26:52.532353 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 21:26:52.620527 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 21:26:52.688146 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 21:26:52.840319 140200711067520 t2t_model.py:2172] Building model body
I0804 21:26:53.533893 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 21:26:54.527805 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 21:26:54.546698 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T21:26:54Z
I0804 21:26:54.713514 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 21:26:54.714270: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:26:54.714716: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 21:26:54.714823: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 21:26:54.714849: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 21:26:54.714871: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 21:26:54.714891: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 21:26:54.714914: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 21:26:54.714936: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 21:26:54.714958: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 21:26:54.715094: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:26:54.715511: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:26:54.715834: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 21:26:54.715875: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 21:26:54.715896: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 21:26:54.715907: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 21:26:54.716197: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:26:54.716611: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:26:54.716955: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 21:26:54.718773 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-208000
I0804 21:26:54.937105 140200711067520 session_manager.py:500] Running local_init_op.
I0804 21:26:54.990136 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 21:27:01.139396 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 21:27:06.541764 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 21:27:11.956675 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 21:27:17.360491 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 21:27:22.701410 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 21:27:28.082128 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 21:27:33.431890 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 21:27:38.824660 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 21:27:44.211163 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 21:27:49.062526 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-21:27:49
I0804 21:27:49.062750 140200711067520 estimator.py:2039] Saving dict for global step 208000: global_step = 208000, loss = 1.1632155, metrics-paper_generation_problem/targets/accuracy = 0.6772446, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8864389, metrics-paper_generation_problem/targets/approx_bleu_score = 0.4942547, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.16325, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5875916, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.69853467
I0804 21:27:49.063206 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 208000: experiment/transformer/transformer_small/output/model.ckpt-208000
I0804 21:27:49.121700 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.66002
I0804 21:27:49.122725 140200711067520 basic_session_run_hooks.py:260] loss = 1.119972, step = 208000 (60.240 sec)
I0804 21:27:52.375464 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7341
I0804 21:27:52.377014 140200711067520 basic_session_run_hooks.py:260] loss = 1.1047922, step = 208100 (3.254 sec)
I0804 21:27:55.563156 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3705
I0804 21:27:55.564710 140200711067520 basic_session_run_hooks.py:260] loss = 1.0596045, step = 208200 (3.188 sec)
I0804 21:27:58.794229 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9495
I0804 21:27:58.795847 140200711067520 basic_session_run_hooks.py:260] loss = 1.0320935, step = 208300 (3.231 sec)
I0804 21:28:02.020396 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9964
I0804 21:28:02.021841 140200711067520 basic_session_run_hooks.py:260] loss = 1.0735085, step = 208400 (3.226 sec)
I0804 21:28:05.234530 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1129
I0804 21:28:05.235708 140200711067520 basic_session_run_hooks.py:260] loss = 1.1495082, step = 208500 (3.214 sec)
I0804 21:28:08.479792 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8138
I0804 21:28:08.481151 140200711067520 basic_session_run_hooks.py:260] loss = 1.0818751, step = 208600 (3.245 sec)
I0804 21:28:11.729447 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7727
I0804 21:28:11.730842 140200711067520 basic_session_run_hooks.py:260] loss = 1.156732, step = 208700 (3.250 sec)
I0804 21:28:15.016897 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4187
I0804 21:28:15.018351 140200711067520 basic_session_run_hooks.py:260] loss = 1.1033305, step = 208800 (3.288 sec)
I0804 21:28:18.285132 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5975
I0804 21:28:18.286659 140200711067520 basic_session_run_hooks.py:260] loss = 1.0016987, step = 208900 (3.268 sec)
I0804 21:28:21.485519 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 209000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:28:21.791481 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:28:21.824442 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.254
I0804 21:28:21.825446 140200711067520 basic_session_run_hooks.py:260] loss = 1.1021962, step = 209000 (3.539 sec)
I0804 21:28:25.049041 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0117
I0804 21:28:25.050386 140200711067520 basic_session_run_hooks.py:260] loss = 1.0309865, step = 209100 (3.225 sec)
I0804 21:28:28.278713 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9628
I0804 21:28:28.280024 140200711067520 basic_session_run_hooks.py:260] loss = 1.0800068, step = 209200 (3.230 sec)
I0804 21:28:31.503404 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0108
I0804 21:28:31.504752 140200711067520 basic_session_run_hooks.py:260] loss = 1.0265377, step = 209300 (3.225 sec)
I0804 21:28:34.729807 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9941
I0804 21:28:34.731229 140200711067520 basic_session_run_hooks.py:260] loss = 1.1222415, step = 209400 (3.226 sec)
I0804 21:28:37.931702 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2315
I0804 21:28:37.933172 140200711067520 basic_session_run_hooks.py:260] loss = 1.0969651, step = 209500 (3.202 sec)
I0804 21:28:41.182126 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7653
I0804 21:28:41.183185 140200711067520 basic_session_run_hooks.py:260] loss = 1.0602196, step = 209600 (3.250 sec)
I0804 21:28:44.342234 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6446
I0804 21:28:44.343665 140200711067520 basic_session_run_hooks.py:260] loss = 1.125179, step = 209700 (3.160 sec)
I0804 21:28:47.470484 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9668
I0804 21:28:47.471809 140200711067520 basic_session_run_hooks.py:260] loss = 1.0555645, step = 209800 (3.128 sec)
I0804 21:28:50.640251 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.548
I0804 21:28:50.641830 140200711067520 basic_session_run_hooks.py:260] loss = 1.1047516, step = 209900 (3.170 sec)
I0804 21:28:53.776792 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 210000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:28:54.073980 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:28:54.114586 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7821
I0804 21:28:54.115586 140200711067520 basic_session_run_hooks.py:260] loss = 1.112896, step = 210000 (3.474 sec)
I0804 21:28:57.270083 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6911
I0804 21:28:57.271403 140200711067520 basic_session_run_hooks.py:260] loss = 1.0779816, step = 210100 (3.156 sec)
I0804 21:29:00.414297 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8044
I0804 21:29:00.415777 140200711067520 basic_session_run_hooks.py:260] loss = 1.0979965, step = 210200 (3.144 sec)
I0804 21:29:03.551107 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8795
I0804 21:29:03.552319 140200711067520 basic_session_run_hooks.py:260] loss = 0.9789487, step = 210300 (3.137 sec)
I0804 21:29:06.756686 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1956
I0804 21:29:06.757977 140200711067520 basic_session_run_hooks.py:260] loss = 1.0572287, step = 210400 (3.206 sec)
I0804 21:29:09.948305 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3319
I0804 21:29:09.949747 140200711067520 basic_session_run_hooks.py:260] loss = 1.0314145, step = 210500 (3.192 sec)
I0804 21:29:13.145572 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.277
I0804 21:29:13.146921 140200711067520 basic_session_run_hooks.py:260] loss = 1.0915885, step = 210600 (3.197 sec)
I0804 21:29:16.350671 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2003
I0804 21:29:16.351908 140200711067520 basic_session_run_hooks.py:260] loss = 1.0652192, step = 210700 (3.205 sec)
I0804 21:29:19.577434 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9908
I0804 21:29:19.578801 140200711067520 basic_session_run_hooks.py:260] loss = 1.1196957, step = 210800 (3.227 sec)
I0804 21:29:22.786070 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1658
I0804 21:29:22.787476 140200711067520 basic_session_run_hooks.py:260] loss = 1.0247225, step = 210900 (3.209 sec)
I0804 21:29:25.979279 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 211000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:29:26.280123 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:29:26.323571 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2685
I0804 21:29:26.324705 140200711067520 basic_session_run_hooks.py:260] loss = 0.96771234, step = 211000 (3.537 sec)
I0804 21:29:29.560559 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8932
I0804 21:29:29.561873 140200711067520 basic_session_run_hooks.py:260] loss = 1.0450896, step = 211100 (3.237 sec)
I0804 21:29:32.800654 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8631
I0804 21:29:32.802184 140200711067520 basic_session_run_hooks.py:260] loss = 1.0983168, step = 211200 (3.240 sec)
I0804 21:29:35.981983 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4335
I0804 21:29:35.983512 140200711067520 basic_session_run_hooks.py:260] loss = 1.0619292, step = 211300 (3.181 sec)
I0804 21:29:39.181079 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2589
I0804 21:29:39.182381 140200711067520 basic_session_run_hooks.py:260] loss = 1.115761, step = 211400 (3.199 sec)
I0804 21:29:42.396391 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1014
I0804 21:29:42.397761 140200711067520 basic_session_run_hooks.py:260] loss = 1.1467235, step = 211500 (3.215 sec)
I0804 21:29:45.588435 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3278
I0804 21:29:45.589610 140200711067520 basic_session_run_hooks.py:260] loss = 1.0474057, step = 211600 (3.192 sec)
I0804 21:29:48.787219 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2617
I0804 21:29:48.789009 140200711067520 basic_session_run_hooks.py:260] loss = 1.0992621, step = 211700 (3.199 sec)
I0804 21:29:51.985344 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2683
I0804 21:29:51.986522 140200711067520 basic_session_run_hooks.py:260] loss = 1.0848857, step = 211800 (3.198 sec)
I0804 21:29:55.203240 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0762
I0804 21:29:55.204714 140200711067520 basic_session_run_hooks.py:260] loss = 1.1071928, step = 211900 (3.218 sec)
I0804 21:29:58.372153 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 212000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:29:58.664120 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:29:58.700849 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5907
I0804 21:29:58.701915 140200711067520 basic_session_run_hooks.py:260] loss = 1.0874542, step = 212000 (3.497 sec)
I0804 21:30:01.853934 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7152
I0804 21:30:01.855003 140200711067520 basic_session_run_hooks.py:260] loss = 1.0270497, step = 212100 (3.153 sec)
I0804 21:30:04.997690 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8091
I0804 21:30:04.998787 140200711067520 basic_session_run_hooks.py:260] loss = 1.0962222, step = 212200 (3.144 sec)
I0804 21:30:08.136051 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8639
I0804 21:30:08.137575 140200711067520 basic_session_run_hooks.py:260] loss = 1.0519813, step = 212300 (3.139 sec)
I0804 21:30:11.305506 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5512
I0804 21:30:11.306578 140200711067520 basic_session_run_hooks.py:260] loss = 1.0602816, step = 212400 (3.169 sec)
I0804 21:30:14.488018 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4217
I0804 21:30:14.489383 140200711067520 basic_session_run_hooks.py:260] loss = 1.1076373, step = 212500 (3.183 sec)
I0804 21:30:17.647109 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6545
I0804 21:30:17.648484 140200711067520 basic_session_run_hooks.py:260] loss = 1.1522998, step = 212600 (3.159 sec)
I0804 21:30:20.809198 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6248
I0804 21:30:20.810656 140200711067520 basic_session_run_hooks.py:260] loss = 1.1196973, step = 212700 (3.162 sec)
I0804 21:30:24.056794 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.792
I0804 21:30:24.058231 140200711067520 basic_session_run_hooks.py:260] loss = 1.0818082, step = 212800 (3.248 sec)
I0804 21:30:27.301619 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8183
I0804 21:30:27.302857 140200711067520 basic_session_run_hooks.py:260] loss = 1.0655849, step = 212900 (3.245 sec)
I0804 21:30:30.473453 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 213000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:30:30.771185 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:30:30.812316 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4841
I0804 21:30:30.813683 140200711067520 basic_session_run_hooks.py:260] loss = 1.0609069, step = 213000 (3.511 sec)
I0804 21:30:34.026218 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1151
I0804 21:30:34.027516 140200711067520 basic_session_run_hooks.py:260] loss = 1.1272929, step = 213100 (3.214 sec)
I0804 21:30:37.241771 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0988
I0804 21:30:37.243051 140200711067520 basic_session_run_hooks.py:260] loss = 1.009708, step = 213200 (3.216 sec)
I0804 21:30:40.434607 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3202
I0804 21:30:40.435985 140200711067520 basic_session_run_hooks.py:260] loss = 0.9985438, step = 213300 (3.193 sec)
I0804 21:30:43.656302 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0397
I0804 21:30:43.657768 140200711067520 basic_session_run_hooks.py:260] loss = 1.0503253, step = 213400 (3.222 sec)
I0804 21:30:46.880657 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0139
I0804 21:30:46.881900 140200711067520 basic_session_run_hooks.py:260] loss = 1.0606813, step = 213500 (3.224 sec)
I0804 21:30:50.112268 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9444
I0804 21:30:50.113726 140200711067520 basic_session_run_hooks.py:260] loss = 1.0621357, step = 213600 (3.232 sec)
I0804 21:30:53.281110 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5572
I0804 21:30:53.282543 140200711067520 basic_session_run_hooks.py:260] loss = 0.9488792, step = 213700 (3.169 sec)
I0804 21:30:56.475340 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3064
I0804 21:30:56.476824 140200711067520 basic_session_run_hooks.py:260] loss = 1.1015611, step = 213800 (3.194 sec)
I0804 21:30:59.636190 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6371
I0804 21:30:59.637540 140200711067520 basic_session_run_hooks.py:260] loss = 1.1580987, step = 213900 (3.161 sec)
I0804 21:31:02.783813 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 214000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:31:03.084144 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:31:03.121989 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6876
I0804 21:31:03.123038 140200711067520 basic_session_run_hooks.py:260] loss = 0.9439693, step = 214000 (3.486 sec)
I0804 21:31:06.312837 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3398
I0804 21:31:06.314149 140200711067520 basic_session_run_hooks.py:260] loss = 1.0808375, step = 214100 (3.191 sec)
I0804 21:31:09.485298 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5213
I0804 21:31:09.486726 140200711067520 basic_session_run_hooks.py:260] loss = 1.0671438, step = 214200 (3.173 sec)
I0804 21:31:12.659576 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5034
I0804 21:31:12.660915 140200711067520 basic_session_run_hooks.py:260] loss = 1.0210564, step = 214300 (3.174 sec)
I0804 21:31:15.880536 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0467
I0804 21:31:15.881943 140200711067520 basic_session_run_hooks.py:260] loss = 1.0103968, step = 214400 (3.221 sec)
I0804 21:31:19.074962 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3043
I0804 21:31:19.076281 140200711067520 basic_session_run_hooks.py:260] loss = 1.0535547, step = 214500 (3.194 sec)
I0804 21:31:22.257919 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4172
I0804 21:31:22.259500 140200711067520 basic_session_run_hooks.py:260] loss = 1.0474252, step = 214600 (3.183 sec)
I0804 21:31:25.457447 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2547
I0804 21:31:25.458747 140200711067520 basic_session_run_hooks.py:260] loss = 1.1534312, step = 214700 (3.199 sec)
I0804 21:31:28.657495 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2497
I0804 21:31:28.658683 140200711067520 basic_session_run_hooks.py:260] loss = 1.003392, step = 214800 (3.200 sec)
I0804 21:31:31.826545 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5554
I0804 21:31:31.827717 140200711067520 basic_session_run_hooks.py:260] loss = 1.1377113, step = 214900 (3.169 sec)
I0804 21:31:34.963446 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 215000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:31:35.264449 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:31:35.304239 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7541
I0804 21:31:35.305321 140200711067520 basic_session_run_hooks.py:260] loss = 1.0620909, step = 215000 (3.478 sec)
I0804 21:31:38.493565 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3552
I0804 21:31:38.494998 140200711067520 basic_session_run_hooks.py:260] loss = 0.98561937, step = 215100 (3.190 sec)
I0804 21:31:41.680433 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3786
I0804 21:31:41.681815 140200711067520 basic_session_run_hooks.py:260] loss = 1.0738376, step = 215200 (3.187 sec)
I0804 21:31:44.898226 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0771
I0804 21:31:44.899372 140200711067520 basic_session_run_hooks.py:260] loss = 1.1894506, step = 215300 (3.218 sec)
I0804 21:31:48.105400 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1802
I0804 21:31:48.106801 140200711067520 basic_session_run_hooks.py:260] loss = 1.0739303, step = 215400 (3.207 sec)
I0804 21:31:51.304633 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2573
I0804 21:31:51.305971 140200711067520 basic_session_run_hooks.py:260] loss = 1.072316, step = 215500 (3.199 sec)
I0804 21:31:54.510840 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1895
I0804 21:31:54.512305 140200711067520 basic_session_run_hooks.py:260] loss = 1.1094962, step = 215600 (3.206 sec)
I0804 21:31:57.718764 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1729
I0804 21:31:57.719998 140200711067520 basic_session_run_hooks.py:260] loss = 1.0879185, step = 215700 (3.208 sec)
I0804 21:32:00.921502 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2234
I0804 21:32:00.922715 140200711067520 basic_session_run_hooks.py:260] loss = 1.0621324, step = 215800 (3.203 sec)
I0804 21:32:04.146826 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0045
I0804 21:32:04.148096 140200711067520 basic_session_run_hooks.py:260] loss = 1.2050617, step = 215900 (3.225 sec)
I0804 21:32:07.348564 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 216000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:32:07.668843 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:32:07.705660 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.0988
I0804 21:32:07.707021 140200711067520 basic_session_run_hooks.py:260] loss = 1.0957549, step = 216000 (3.559 sec)
I0804 21:32:10.956106 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7653
I0804 21:32:10.957246 140200711067520 basic_session_run_hooks.py:260] loss = 1.1561966, step = 216100 (3.250 sec)
I0804 21:32:14.176576 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0515
I0804 21:32:14.177869 140200711067520 basic_session_run_hooks.py:260] loss = 1.0422009, step = 216200 (3.221 sec)
I0804 21:32:17.358371 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4287
I0804 21:32:17.359841 140200711067520 basic_session_run_hooks.py:260] loss = 1.0427254, step = 216300 (3.182 sec)
I0804 21:32:20.541387 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4166
I0804 21:32:20.542830 140200711067520 basic_session_run_hooks.py:260] loss = 1.0469387, step = 216400 (3.183 sec)
I0804 21:32:23.716588 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4944
I0804 21:32:23.717665 140200711067520 basic_session_run_hooks.py:260] loss = 1.1565726, step = 216500 (3.175 sec)
I0804 21:32:26.930943 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1102
I0804 21:32:26.932157 140200711067520 basic_session_run_hooks.py:260] loss = 0.9452041, step = 216600 (3.214 sec)
I0804 21:32:30.111652 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4394
I0804 21:32:30.112744 140200711067520 basic_session_run_hooks.py:260] loss = 1.0280825, step = 216700 (3.181 sec)
I0804 21:32:33.261917 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7435
I0804 21:32:33.263575 140200711067520 basic_session_run_hooks.py:260] loss = 1.0245703, step = 216800 (3.151 sec)
I0804 21:32:36.397473 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8923
I0804 21:32:36.398772 140200711067520 basic_session_run_hooks.py:260] loss = 1.0821853, step = 216900 (3.135 sec)
I0804 21:32:39.490575 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 217000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:32:39.788402 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:32:39.830902 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.1251
I0804 21:32:39.831987 140200711067520 basic_session_run_hooks.py:260] loss = 0.99919194, step = 217000 (3.433 sec)
I0804 21:32:43.080310 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.775
I0804 21:32:43.081701 140200711067520 basic_session_run_hooks.py:260] loss = 1.0866444, step = 217100 (3.250 sec)
I0804 21:32:46.261224 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4376
I0804 21:32:46.262771 140200711067520 basic_session_run_hooks.py:260] loss = 1.1221713, step = 217200 (3.181 sec)
I0804 21:32:49.438490 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4739
I0804 21:32:49.439822 140200711067520 basic_session_run_hooks.py:260] loss = 1.0402652, step = 217300 (3.177 sec)
I0804 21:32:52.583650 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7945
I0804 21:32:52.584986 140200711067520 basic_session_run_hooks.py:260] loss = 1.1212548, step = 217400 (3.145 sec)
I0804 21:32:55.750571 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5768
I0804 21:32:55.752244 140200711067520 basic_session_run_hooks.py:260] loss = 1.1150932, step = 217500 (3.167 sec)
I0804 21:32:58.891660 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.836
I0804 21:32:58.893235 140200711067520 basic_session_run_hooks.py:260] loss = 1.0800558, step = 217600 (3.141 sec)
I0804 21:33:02.036540 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7979
I0804 21:33:02.038184 140200711067520 basic_session_run_hooks.py:260] loss = 1.0737025, step = 217700 (3.145 sec)
I0804 21:33:05.186833 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7428
I0804 21:33:05.188227 140200711067520 basic_session_run_hooks.py:260] loss = 1.1170472, step = 217800 (3.150 sec)
I0804 21:33:08.382198 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2953
I0804 21:33:08.383639 140200711067520 basic_session_run_hooks.py:260] loss = 1.0862323, step = 217900 (3.195 sec)
I0804 21:33:11.569524 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 218000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:33:11.871132 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:33:11.922072 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2495
I0804 21:33:11.923299 140200711067520 basic_session_run_hooks.py:260] loss = 1.1256591, step = 218000 (3.540 sec)
I0804 21:33:15.138221 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0931
I0804 21:33:15.139678 140200711067520 basic_session_run_hooks.py:260] loss = 1.0554107, step = 218100 (3.216 sec)
I0804 21:33:18.331981 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3112
I0804 21:33:18.333115 140200711067520 basic_session_run_hooks.py:260] loss = 1.1074111, step = 218200 (3.193 sec)
I0804 21:33:21.525080 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3177
I0804 21:33:21.526511 140200711067520 basic_session_run_hooks.py:260] loss = 1.0109075, step = 218300 (3.193 sec)
I0804 21:33:24.701768 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4793
I0804 21:33:24.703037 140200711067520 basic_session_run_hooks.py:260] loss = 1.0735339, step = 218400 (3.177 sec)
I0804 21:33:27.929226 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.984
I0804 21:33:27.930343 140200711067520 basic_session_run_hooks.py:260] loss = 1.1045538, step = 218500 (3.227 sec)
I0804 21:33:31.118873 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3515
I0804 21:33:31.120348 140200711067520 basic_session_run_hooks.py:260] loss = 1.1131818, step = 218600 (3.190 sec)
I0804 21:33:34.364785 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.808
I0804 21:33:34.366103 140200711067520 basic_session_run_hooks.py:260] loss = 1.1755759, step = 218700 (3.246 sec)
I0804 21:33:37.593308 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9738
I0804 21:33:37.594781 140200711067520 basic_session_run_hooks.py:260] loss = 1.1100144, step = 218800 (3.229 sec)
I0804 21:33:40.819835 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9932
I0804 21:33:40.821270 140200711067520 basic_session_run_hooks.py:260] loss = 1.117846, step = 218900 (3.226 sec)
I0804 21:33:44.021180 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 219000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:33:44.323913 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:33:44.364271 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.213
I0804 21:33:44.365203 140200711067520 basic_session_run_hooks.py:260] loss = 1.0980419, step = 219000 (3.544 sec)
I0804 21:33:47.604389 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8634
I0804 21:33:47.605815 140200711067520 basic_session_run_hooks.py:260] loss = 1.0444968, step = 219100 (3.241 sec)
I0804 21:33:50.838175 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9234
I0804 21:33:50.839480 140200711067520 basic_session_run_hooks.py:260] loss = 1.0977076, step = 219200 (3.234 sec)
I0804 21:33:54.060919 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0296
I0804 21:33:54.062175 140200711067520 basic_session_run_hooks.py:260] loss = 1.1180186, step = 219300 (3.223 sec)
I0804 21:33:57.316652 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7151
I0804 21:33:57.317821 140200711067520 basic_session_run_hooks.py:260] loss = 1.0498248, step = 219400 (3.256 sec)
I0804 21:34:00.557660 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8545
I0804 21:34:00.559002 140200711067520 basic_session_run_hooks.py:260] loss = 1.0171392, step = 219500 (3.241 sec)
I0804 21:34:03.714514 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6772
I0804 21:34:03.715955 140200711067520 basic_session_run_hooks.py:260] loss = 1.0394205, step = 219600 (3.157 sec)
I0804 21:34:06.894639 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4451
I0804 21:34:06.896183 140200711067520 basic_session_run_hooks.py:260] loss = 1.0645446, step = 219700 (3.180 sec)
I0804 21:34:10.060574 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5865
I0804 21:34:10.062147 140200711067520 basic_session_run_hooks.py:260] loss = 1.0614727, step = 219800 (3.166 sec)
I0804 21:34:13.218270 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6687
I0804 21:34:13.219664 140200711067520 basic_session_run_hooks.py:260] loss = 1.0990589, step = 219900 (3.158 sec)
I0804 21:34:16.362264 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 220000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:34:16.671204 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:34:16.713469 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6105
I0804 21:34:16.714467 140200711067520 basic_session_run_hooks.py:260] loss = 0.9845242, step = 220000 (3.495 sec)
I0804 21:34:19.875834 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6219
I0804 21:34:19.877131 140200711067520 basic_session_run_hooks.py:260] loss = 1.073499, step = 220100 (3.163 sec)
I0804 21:34:23.024921 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7552
I0804 21:34:23.026248 140200711067520 basic_session_run_hooks.py:260] loss = 1.1392233, step = 220200 (3.149 sec)
I0804 21:34:26.218338 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3145
I0804 21:34:26.219711 140200711067520 basic_session_run_hooks.py:260] loss = 1.0618665, step = 220300 (3.193 sec)
I0804 21:34:29.383329 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5956
I0804 21:34:29.384763 140200711067520 basic_session_run_hooks.py:260] loss = 1.115074, step = 220400 (3.165 sec)
I0804 21:34:32.538407 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6949
I0804 21:34:32.539808 140200711067520 basic_session_run_hooks.py:260] loss = 1.0031439, step = 220500 (3.155 sec)
I0804 21:34:35.701464 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6151
I0804 21:34:35.702906 140200711067520 basic_session_run_hooks.py:260] loss = 1.0504954, step = 220600 (3.163 sec)
I0804 21:34:38.861980 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6403
I0804 21:34:38.863434 140200711067520 basic_session_run_hooks.py:260] loss = 1.0386574, step = 220700 (3.161 sec)
I0804 21:34:42.024827 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.617
I0804 21:34:42.026117 140200711067520 basic_session_run_hooks.py:260] loss = 1.0272084, step = 220800 (3.163 sec)
I0804 21:34:45.192012 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5739
I0804 21:34:45.193534 140200711067520 basic_session_run_hooks.py:260] loss = 1.0692806, step = 220900 (3.167 sec)
I0804 21:34:48.354452 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 221000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:34:48.648157 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:34:48.690867 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5806
I0804 21:34:48.691941 140200711067520 basic_session_run_hooks.py:260] loss = 1.0665172, step = 221000 (3.498 sec)
I0804 21:34:51.930593 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8673
I0804 21:34:51.932125 140200711067520 basic_session_run_hooks.py:260] loss = 1.0096669, step = 221100 (3.240 sec)
I0804 21:34:55.187339 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7054
I0804 21:34:55.188449 140200711067520 basic_session_run_hooks.py:260] loss = 1.0062926, step = 221200 (3.256 sec)
I0804 21:34:58.467450 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4868
I0804 21:34:58.468667 140200711067520 basic_session_run_hooks.py:260] loss = 1.0550654, step = 221300 (3.280 sec)
I0804 21:35:01.733979 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6133
I0804 21:35:01.735142 140200711067520 basic_session_run_hooks.py:260] loss = 1.0578343, step = 221400 (3.266 sec)
I0804 21:35:04.965388 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9462
I0804 21:35:04.966827 140200711067520 basic_session_run_hooks.py:260] loss = 1.1090732, step = 221500 (3.232 sec)
I0804 21:35:08.193613 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9768
I0804 21:35:08.194782 140200711067520 basic_session_run_hooks.py:260] loss = 1.0937297, step = 221600 (3.228 sec)
I0804 21:35:11.438388 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8188
I0804 21:35:11.439878 140200711067520 basic_session_run_hooks.py:260] loss = 1.0372883, step = 221700 (3.245 sec)
I0804 21:35:14.706675 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.5971
I0804 21:35:14.707973 140200711067520 basic_session_run_hooks.py:260] loss = 0.9393958, step = 221800 (3.268 sec)
I0804 21:35:17.939704 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9307
I0804 21:35:17.940799 140200711067520 basic_session_run_hooks.py:260] loss = 1.0968131, step = 221900 (3.233 sec)
I0804 21:35:21.060581 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 222000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:35:21.366719 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:35:21.402782 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8759
I0804 21:35:21.403769 140200711067520 basic_session_run_hooks.py:260] loss = 1.1306627, step = 222000 (3.463 sec)
I0804 21:35:24.547210 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8025
I0804 21:35:24.548693 140200711067520 basic_session_run_hooks.py:260] loss = 1.0109096, step = 222100 (3.145 sec)
I0804 21:35:27.704639 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6715
I0804 21:35:27.706010 140200711067520 basic_session_run_hooks.py:260] loss = 1.1559556, step = 222200 (3.157 sec)
I0804 21:35:30.880089 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4916
I0804 21:35:30.881236 140200711067520 basic_session_run_hooks.py:260] loss = 1.0636022, step = 222300 (3.175 sec)
I0804 21:35:34.019044 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8578
I0804 21:35:34.020484 140200711067520 basic_session_run_hooks.py:260] loss = 1.0687085, step = 222400 (3.139 sec)
I0804 21:35:37.172576 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7107
I0804 21:35:37.173904 140200711067520 basic_session_run_hooks.py:260] loss = 1.0507942, step = 222500 (3.153 sec)
I0804 21:35:40.328183 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6894
I0804 21:35:40.329647 140200711067520 basic_session_run_hooks.py:260] loss = 1.108968, step = 222600 (3.156 sec)
I0804 21:35:43.530199 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2302
I0804 21:35:43.531771 140200711067520 basic_session_run_hooks.py:260] loss = 1.0217676, step = 222700 (3.202 sec)
I0804 21:35:46.699893 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5489
I0804 21:35:46.701531 140200711067520 basic_session_run_hooks.py:260] loss = 1.0905676, step = 222800 (3.170 sec)
I0804 21:35:49.901813 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2312
I0804 21:35:49.903061 140200711067520 basic_session_run_hooks.py:260] loss = 1.0607625, step = 222900 (3.202 sec)
I0804 21:35:53.032622 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 223000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:35:53.327353 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:35:53.367242 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8562
I0804 21:35:53.368304 140200711067520 basic_session_run_hooks.py:260] loss = 1.0075176, step = 223000 (3.465 sec)
I0804 21:35:56.567482 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2481
I0804 21:35:56.568661 140200711067520 basic_session_run_hooks.py:260] loss = 1.1599233, step = 223100 (3.200 sec)
I0804 21:35:59.740089 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5198
I0804 21:35:59.741513 140200711067520 basic_session_run_hooks.py:260] loss = 1.0649606, step = 223200 (3.173 sec)
I0804 21:36:02.907191 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5743
I0804 21:36:02.908358 140200711067520 basic_session_run_hooks.py:260] loss = 1.0522668, step = 223300 (3.167 sec)
I0804 21:36:06.073328 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5847
I0804 21:36:06.074747 140200711067520 basic_session_run_hooks.py:260] loss = 1.0550487, step = 223400 (3.166 sec)
I0804 21:36:09.339740 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.6142
I0804 21:36:09.341344 140200711067520 basic_session_run_hooks.py:260] loss = 1.1103789, step = 223500 (3.267 sec)
I0804 21:36:12.580230 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8597
I0804 21:36:12.581752 140200711067520 basic_session_run_hooks.py:260] loss = 1.121515, step = 223600 (3.240 sec)
I0804 21:36:15.827195 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7979
I0804 21:36:15.828703 140200711067520 basic_session_run_hooks.py:260] loss = 0.99285054, step = 223700 (3.247 sec)
I0804 21:36:19.071115 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.827
I0804 21:36:19.072510 140200711067520 basic_session_run_hooks.py:260] loss = 1.0404335, step = 223800 (3.244 sec)
I0804 21:36:22.306205 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.911
I0804 21:36:22.307644 140200711067520 basic_session_run_hooks.py:260] loss = 1.0605198, step = 223900 (3.235 sec)
I0804 21:36:25.516498 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 224000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:36:25.821990 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:36:25.858863 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.1478
I0804 21:36:25.859924 140200711067520 basic_session_run_hooks.py:260] loss = 1.0638331, step = 224000 (3.552 sec)
I0804 21:36:29.106234 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7944
I0804 21:36:29.107406 140200711067520 basic_session_run_hooks.py:260] loss = 1.0377556, step = 224100 (3.247 sec)
I0804 21:36:32.348674 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8408
I0804 21:36:32.350018 140200711067520 basic_session_run_hooks.py:260] loss = 1.1124936, step = 224200 (3.243 sec)
I0804 21:36:35.573818 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0065
I0804 21:36:35.575115 140200711067520 basic_session_run_hooks.py:260] loss = 1.0980145, step = 224300 (3.225 sec)
I0804 21:36:38.752272 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.462
I0804 21:36:38.753720 140200711067520 basic_session_run_hooks.py:260] loss = 1.077118, step = 224400 (3.179 sec)
I0804 21:36:41.916196 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6061
I0804 21:36:41.918099 140200711067520 basic_session_run_hooks.py:260] loss = 1.0585124, step = 224500 (3.164 sec)
I0804 21:36:45.075251 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6551
I0804 21:36:45.076845 140200711067520 basic_session_run_hooks.py:260] loss = 1.0607466, step = 224600 (3.159 sec)
I0804 21:36:48.246672 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5316
I0804 21:36:48.248111 140200711067520 basic_session_run_hooks.py:260] loss = 1.0254956, step = 224700 (3.171 sec)
I0804 21:36:51.389965 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8137
I0804 21:36:51.391455 140200711067520 basic_session_run_hooks.py:260] loss = 1.0540202, step = 224800 (3.143 sec)
I0804 21:36:54.537089 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7753
I0804 21:36:54.538624 140200711067520 basic_session_run_hooks.py:260] loss = 1.018664, step = 224900 (3.147 sec)
I0804 21:36:57.668853 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 225000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:36:57.973995 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 21:36:57.975388 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 21:36:58.120602 140200711067520 estimator.py:1145] Calling model_fn.
I0804 21:36:58.121630 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 21:36:58.122027 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 21:36:58.122117 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 21:36:58.122200 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 21:36:58.122267 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 21:36:58.122348 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 21:36:58.122415 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 21:36:58.209110 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 21:36:58.267072 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 21:36:58.404612 140200711067520 t2t_model.py:2172] Building model body
I0804 21:36:59.351598 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 21:37:00.059998 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 21:37:00.080130 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T21:37:00Z
I0804 21:37:00.243007 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 21:37:00.243623: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:37:00.244017: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 21:37:00.244109: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 21:37:00.244136: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 21:37:00.244157: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 21:37:00.244180: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 21:37:00.244199: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 21:37:00.244218: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 21:37:00.244238: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 21:37:00.244342: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:37:00.244802: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:37:00.245121: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 21:37:00.245161: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 21:37:00.245174: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 21:37:00.245184: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 21:37:00.245474: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:37:00.245877: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:37:00.246236: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 21:37:00.247739 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-225000
I0804 21:37:00.455331 140200711067520 session_manager.py:500] Running local_init_op.
I0804 21:37:00.500869 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 21:37:06.740935 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 21:37:12.091328 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 21:37:17.521615 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 21:37:22.906087 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 21:37:28.273287 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 21:37:33.635584 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 21:37:38.986086 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 21:37:44.356673 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 21:37:49.705099 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 21:37:54.616703 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-21:37:54
I0804 21:37:54.616927 140200711067520 estimator.py:2039] Saving dict for global step 225000: global_step = 225000, loss = 1.1603231, metrics-paper_generation_problem/targets/accuracy = 0.67825276, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.88668895, metrics-paper_generation_problem/targets/approx_bleu_score = 0.4954981, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1603596, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.58919674, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.69938874
I0804 21:37:54.617377 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 225000: experiment/transformer/transformer_small/output/model.ckpt-225000
I0804 21:37:54.671779 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.66293
I0804 21:37:54.673111 140200711067520 basic_session_run_hooks.py:260] loss = 1.0412966, step = 225000 (60.134 sec)
I0804 21:37:57.919399 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7921
I0804 21:37:57.920792 140200711067520 basic_session_run_hooks.py:260] loss = 1.0747746, step = 225100 (3.248 sec)
I0804 21:38:01.112833 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3144
I0804 21:38:01.114253 140200711067520 basic_session_run_hooks.py:260] loss = 1.030618, step = 225200 (3.193 sec)
I0804 21:38:04.312991 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2484
I0804 21:38:04.314345 140200711067520 basic_session_run_hooks.py:260] loss = 1.1233718, step = 225300 (3.200 sec)
I0804 21:38:07.534989 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0366
I0804 21:38:07.536501 140200711067520 basic_session_run_hooks.py:260] loss = 1.0498263, step = 225400 (3.222 sec)
I0804 21:38:10.757383 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0331
I0804 21:38:10.758515 140200711067520 basic_session_run_hooks.py:260] loss = 0.98746026, step = 225500 (3.222 sec)
I0804 21:38:14.004359 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7975
I0804 21:38:14.005671 140200711067520 basic_session_run_hooks.py:260] loss = 1.1017932, step = 225600 (3.247 sec)
I0804 21:38:17.246465 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8444
I0804 21:38:17.247732 140200711067520 basic_session_run_hooks.py:260] loss = 1.0178926, step = 225700 (3.242 sec)
I0804 21:38:20.483771 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8898
I0804 21:38:20.485177 140200711067520 basic_session_run_hooks.py:260] loss = 1.0664741, step = 225800 (3.237 sec)
I0804 21:38:23.773575 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.3971
I0804 21:38:23.775268 140200711067520 basic_session_run_hooks.py:260] loss = 1.075894, step = 225900 (3.290 sec)
I0804 21:38:27.024256 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 226000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:38:27.329359 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:38:27.373456 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 27.7785
I0804 21:38:27.374636 140200711067520 basic_session_run_hooks.py:260] loss = 1.1009996, step = 226000 (3.599 sec)
I0804 21:38:30.629258 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7144
I0804 21:38:30.630750 140200711067520 basic_session_run_hooks.py:260] loss = 1.0368599, step = 226100 (3.256 sec)
I0804 21:38:33.858433 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9679
I0804 21:38:33.859796 140200711067520 basic_session_run_hooks.py:260] loss = 1.0550992, step = 226200 (3.229 sec)
I0804 21:38:37.068322 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1536
I0804 21:38:37.069738 140200711067520 basic_session_run_hooks.py:260] loss = 0.9654957, step = 226300 (3.210 sec)
I0804 21:38:40.316176 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7895
I0804 21:38:40.317521 140200711067520 basic_session_run_hooks.py:260] loss = 1.0024687, step = 226400 (3.248 sec)
I0804 21:38:43.546176 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9597
I0804 21:38:43.547502 140200711067520 basic_session_run_hooks.py:260] loss = 1.0229592, step = 226500 (3.230 sec)
I0804 21:38:46.753021 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1835
I0804 21:38:46.754550 140200711067520 basic_session_run_hooks.py:260] loss = 1.1276205, step = 226600 (3.207 sec)
I0804 21:38:49.923189 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5441
I0804 21:38:49.924715 140200711067520 basic_session_run_hooks.py:260] loss = 1.027961, step = 226700 (3.170 sec)
I0804 21:38:53.101408 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.464
I0804 21:38:53.102945 140200711067520 basic_session_run_hooks.py:260] loss = 1.0351074, step = 226800 (3.178 sec)
I0804 21:38:56.293028 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3321
I0804 21:38:56.294504 140200711067520 basic_session_run_hooks.py:260] loss = 1.0202962, step = 226900 (3.192 sec)
I0804 21:38:59.562267 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 227000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:38:59.863342 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:38:59.908333 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 27.6599
I0804 21:38:59.909309 140200711067520 basic_session_run_hooks.py:260] loss = 1.1464305, step = 227000 (3.615 sec)
I0804 21:39:03.104080 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2919
I0804 21:39:03.105569 140200711067520 basic_session_run_hooks.py:260] loss = 1.0552921, step = 227100 (3.196 sec)
I0804 21:39:06.305931 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2319
I0804 21:39:06.307254 140200711067520 basic_session_run_hooks.py:260] loss = 1.0073179, step = 227200 (3.202 sec)
I0804 21:39:09.513444 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.177
I0804 21:39:09.514811 140200711067520 basic_session_run_hooks.py:260] loss = 0.9870765, step = 227300 (3.208 sec)
I0804 21:39:12.721231 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1739
I0804 21:39:12.722766 140200711067520 basic_session_run_hooks.py:260] loss = 1.1089511, step = 227400 (3.208 sec)
I0804 21:39:15.948914 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9835
I0804 21:39:15.950088 140200711067520 basic_session_run_hooks.py:260] loss = 1.0985713, step = 227500 (3.227 sec)
I0804 21:39:19.161293 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.128
I0804 21:39:19.162899 140200711067520 basic_session_run_hooks.py:260] loss = 1.0555142, step = 227600 (3.213 sec)
I0804 21:39:22.382606 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0432
I0804 21:39:22.383773 140200711067520 basic_session_run_hooks.py:260] loss = 1.1311486, step = 227700 (3.221 sec)
I0804 21:39:25.611279 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9725
I0804 21:39:25.612731 140200711067520 basic_session_run_hooks.py:260] loss = 1.1239816, step = 227800 (3.229 sec)
I0804 21:39:28.816586 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1985
I0804 21:39:28.817854 140200711067520 basic_session_run_hooks.py:260] loss = 0.9361052, step = 227900 (3.205 sec)
I0804 21:39:31.962359 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 228000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:39:32.266844 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:39:32.301978 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6908
I0804 21:39:32.303231 140200711067520 basic_session_run_hooks.py:260] loss = 1.1122743, step = 228000 (3.485 sec)
I0804 21:39:35.483943 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4273
I0804 21:39:35.485393 140200711067520 basic_session_run_hooks.py:260] loss = 1.117333, step = 228100 (3.182 sec)
I0804 21:39:38.674948 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3381
I0804 21:39:38.676356 140200711067520 basic_session_run_hooks.py:260] loss = 1.13173, step = 228200 (3.191 sec)
I0804 21:39:41.847917 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5161
I0804 21:39:41.849412 140200711067520 basic_session_run_hooks.py:260] loss = 1.1103963, step = 228300 (3.173 sec)
I0804 21:39:45.018387 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.541
I0804 21:39:45.019944 140200711067520 basic_session_run_hooks.py:260] loss = 1.0982221, step = 228400 (3.171 sec)
I0804 21:39:48.209022 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3419
I0804 21:39:48.210457 140200711067520 basic_session_run_hooks.py:260] loss = 1.0509295, step = 228500 (3.190 sec)
I0804 21:39:51.426283 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0823
I0804 21:39:51.427880 140200711067520 basic_session_run_hooks.py:260] loss = 1.1004397, step = 228600 (3.217 sec)
I0804 21:39:54.633787 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1768
I0804 21:39:54.635096 140200711067520 basic_session_run_hooks.py:260] loss = 1.1212882, step = 228700 (3.207 sec)
I0804 21:39:57.841734 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1726
I0804 21:39:57.843210 140200711067520 basic_session_run_hooks.py:260] loss = 1.0423591, step = 228800 (3.208 sec)
I0804 21:40:01.056014 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1113
I0804 21:40:01.057132 140200711067520 basic_session_run_hooks.py:260] loss = 1.0979847, step = 228900 (3.214 sec)
I0804 21:40:04.257563 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 229000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:40:04.567340 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:40:04.604076 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.1841
I0804 21:40:04.605121 140200711067520 basic_session_run_hooks.py:260] loss = 1.0505002, step = 229000 (3.548 sec)
I0804 21:40:07.816784 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1267
I0804 21:40:07.818211 140200711067520 basic_session_run_hooks.py:260] loss = 0.99451846, step = 229100 (3.213 sec)
I0804 21:40:11.020249 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2162
I0804 21:40:11.021636 140200711067520 basic_session_run_hooks.py:260] loss = 1.0549765, step = 229200 (3.203 sec)
I0804 21:40:14.206743 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3824
I0804 21:40:14.207793 140200711067520 basic_session_run_hooks.py:260] loss = 1.0166072, step = 229300 (3.186 sec)
I0804 21:40:17.444475 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8859
I0804 21:40:17.445869 140200711067520 basic_session_run_hooks.py:260] loss = 1.0773181, step = 229400 (3.238 sec)
I0804 21:40:20.612874 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5615
I0804 21:40:20.614355 140200711067520 basic_session_run_hooks.py:260] loss = 1.0859579, step = 229500 (3.168 sec)
I0804 21:40:23.791314 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.462
I0804 21:40:23.793037 140200711067520 basic_session_run_hooks.py:260] loss = 1.0078558, step = 229600 (3.179 sec)
I0804 21:40:26.978708 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3737
I0804 21:40:26.980032 140200711067520 basic_session_run_hooks.py:260] loss = 1.0984629, step = 229700 (3.187 sec)
I0804 21:40:30.151231 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5206
I0804 21:40:30.152699 140200711067520 basic_session_run_hooks.py:260] loss = 1.1522447, step = 229800 (3.173 sec)
I0804 21:40:33.367125 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0956
I0804 21:40:33.368706 140200711067520 basic_session_run_hooks.py:260] loss = 1.0111748, step = 229900 (3.216 sec)
I0804 21:40:36.520762 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 230000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:40:36.828833 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:40:36.872705 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5258
I0804 21:40:36.873844 140200711067520 basic_session_run_hooks.py:260] loss = 1.0309488, step = 230000 (3.505 sec)
I0804 21:40:40.065150 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3241
I0804 21:40:40.066606 140200711067520 basic_session_run_hooks.py:260] loss = 1.1470646, step = 230100 (3.193 sec)
I0804 21:40:43.271497 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1882
I0804 21:40:43.272720 140200711067520 basic_session_run_hooks.py:260] loss = 1.1304742, step = 230200 (3.206 sec)
I0804 21:40:46.426956 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6911
I0804 21:40:46.428337 140200711067520 basic_session_run_hooks.py:260] loss = 0.9931978, step = 230300 (3.156 sec)
I0804 21:40:49.601093 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5046
I0804 21:40:49.602414 140200711067520 basic_session_run_hooks.py:260] loss = 1.0967582, step = 230400 (3.174 sec)
I0804 21:40:52.746317 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7942
I0804 21:40:52.747795 140200711067520 basic_session_run_hooks.py:260] loss = 1.0801775, step = 230500 (3.145 sec)
I0804 21:40:55.885215 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8583
I0804 21:40:55.886781 140200711067520 basic_session_run_hooks.py:260] loss = 1.0554054, step = 230600 (3.139 sec)
I0804 21:40:59.039753 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7004
I0804 21:40:59.040981 140200711067520 basic_session_run_hooks.py:260] loss = 1.0356518, step = 230700 (3.154 sec)
I0804 21:41:02.197182 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6713
I0804 21:41:02.198773 140200711067520 basic_session_run_hooks.py:260] loss = 1.1068004, step = 230800 (3.158 sec)
I0804 21:41:05.376465 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4539
I0804 21:41:05.377875 140200711067520 basic_session_run_hooks.py:260] loss = 1.0569111, step = 230900 (3.179 sec)
I0804 21:41:08.573752 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 231000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:41:08.870705 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:41:08.913198 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2742
I0804 21:41:08.914264 140200711067520 basic_session_run_hooks.py:260] loss = 1.084612, step = 231000 (3.536 sec)
I0804 21:41:12.146030 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9331
I0804 21:41:12.147362 140200711067520 basic_session_run_hooks.py:260] loss = 1.0633703, step = 231100 (3.233 sec)
I0804 21:41:15.373292 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9859
I0804 21:41:15.374706 140200711067520 basic_session_run_hooks.py:260] loss = 1.0458615, step = 231200 (3.227 sec)
I0804 21:41:18.596484 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0253
I0804 21:41:18.597884 140200711067520 basic_session_run_hooks.py:260] loss = 0.99701774, step = 231300 (3.223 sec)
I0804 21:41:21.821294 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0093
I0804 21:41:21.822814 140200711067520 basic_session_run_hooks.py:260] loss = 1.0710945, step = 231400 (3.225 sec)
I0804 21:41:25.045784 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0127
I0804 21:41:25.047168 140200711067520 basic_session_run_hooks.py:260] loss = 1.0106586, step = 231500 (3.224 sec)
I0804 21:41:28.275327 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9642
I0804 21:41:28.276635 140200711067520 basic_session_run_hooks.py:260] loss = 1.0692348, step = 231600 (3.229 sec)
I0804 21:41:31.496262 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0469
I0804 21:41:31.497736 140200711067520 basic_session_run_hooks.py:260] loss = 1.0110937, step = 231700 (3.221 sec)
I0804 21:41:34.752546 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.71
I0804 21:41:34.753901 140200711067520 basic_session_run_hooks.py:260] loss = 1.1160303, step = 231800 (3.256 sec)
I0804 21:41:37.924916 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5219
I0804 21:41:37.926051 140200711067520 basic_session_run_hooks.py:260] loss = 0.9783343, step = 231900 (3.172 sec)
I0804 21:41:41.042064 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 232000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:41:41.344769 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:41:41.386350 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.8896
I0804 21:41:41.387320 140200711067520 basic_session_run_hooks.py:260] loss = 1.0258989, step = 232000 (3.461 sec)
I0804 21:41:44.571481 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3963
I0804 21:41:44.572515 140200711067520 basic_session_run_hooks.py:260] loss = 1.0491873, step = 232100 (3.185 sec)
I0804 21:41:47.734115 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6191
I0804 21:41:47.735476 140200711067520 basic_session_run_hooks.py:260] loss = 1.0997097, step = 232200 (3.163 sec)
I0804 21:41:50.865285 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9369
I0804 21:41:50.866634 140200711067520 basic_session_run_hooks.py:260] loss = 1.1061924, step = 232300 (3.131 sec)
I0804 21:41:54.048371 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.416
I0804 21:41:54.049695 140200711067520 basic_session_run_hooks.py:260] loss = 1.0338413, step = 232400 (3.183 sec)
I0804 21:41:57.216307 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5662
I0804 21:41:57.217677 140200711067520 basic_session_run_hooks.py:260] loss = 1.0680993, step = 232500 (3.168 sec)
I0804 21:42:00.429515 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1219
I0804 21:42:00.431056 140200711067520 basic_session_run_hooks.py:260] loss = 1.1310732, step = 232600 (3.213 sec)
I0804 21:42:03.587292 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6676
I0804 21:42:03.589077 140200711067520 basic_session_run_hooks.py:260] loss = 0.9977112, step = 232700 (3.158 sec)
I0804 21:42:06.742449 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6942
I0804 21:42:06.743656 140200711067520 basic_session_run_hooks.py:260] loss = 1.0966942, step = 232800 (3.155 sec)
I0804 21:42:09.906820 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6018
I0804 21:42:09.907960 140200711067520 basic_session_run_hooks.py:260] loss = 1.0769311, step = 232900 (3.164 sec)
I0804 21:42:13.049041 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 233000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:42:13.343145 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:42:13.385301 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7481
I0804 21:42:13.386391 140200711067520 basic_session_run_hooks.py:260] loss = 1.0484529, step = 233000 (3.478 sec)
I0804 21:42:16.539087 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7082
I0804 21:42:16.540495 140200711067520 basic_session_run_hooks.py:260] loss = 1.0306649, step = 233100 (3.154 sec)
I0804 21:42:19.682455 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8131
I0804 21:42:19.683732 140200711067520 basic_session_run_hooks.py:260] loss = 0.9541991, step = 233200 (3.143 sec)
I0804 21:42:22.832933 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7409
I0804 21:42:22.833998 140200711067520 basic_session_run_hooks.py:260] loss = 1.077165, step = 233300 (3.150 sec)
I0804 21:42:26.116300 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.4566
I0804 21:42:26.117819 140200711067520 basic_session_run_hooks.py:260] loss = 1.098256, step = 233400 (3.284 sec)
I0804 21:42:29.348315 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9403
I0804 21:42:29.349690 140200711067520 basic_session_run_hooks.py:260] loss = 1.1523637, step = 233500 (3.232 sec)
I0804 21:42:32.589391 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8541
I0804 21:42:32.590638 140200711067520 basic_session_run_hooks.py:260] loss = 1.0876305, step = 233600 (3.241 sec)
I0804 21:42:35.810154 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0486
I0804 21:42:35.811767 140200711067520 basic_session_run_hooks.py:260] loss = 1.0063478, step = 233700 (3.221 sec)
I0804 21:42:39.030603 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0518
I0804 21:42:39.032034 140200711067520 basic_session_run_hooks.py:260] loss = 1.1077042, step = 233800 (3.220 sec)
I0804 21:42:42.238704 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1708
I0804 21:42:42.239988 140200711067520 basic_session_run_hooks.py:260] loss = 1.0480468, step = 233900 (3.208 sec)
I0804 21:42:45.420908 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 234000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:42:45.713332 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:42:45.755212 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4371
I0804 21:42:45.756446 140200711067520 basic_session_run_hooks.py:260] loss = 1.1343548, step = 234000 (3.516 sec)
I0804 21:42:48.957231 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2305
I0804 21:42:48.958486 140200711067520 basic_session_run_hooks.py:260] loss = 1.0245591, step = 234100 (3.202 sec)
I0804 21:42:52.178557 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0433
I0804 21:42:52.179984 140200711067520 basic_session_run_hooks.py:260] loss = 1.1224388, step = 234200 (3.222 sec)
I0804 21:42:55.337349 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6574
I0804 21:42:55.338848 140200711067520 basic_session_run_hooks.py:260] loss = 0.991715, step = 234300 (3.159 sec)
I0804 21:42:58.537179 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2518
I0804 21:42:58.538739 140200711067520 basic_session_run_hooks.py:260] loss = 1.1411386, step = 234400 (3.200 sec)
I0804 21:43:01.692112 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6963
I0804 21:43:01.693543 140200711067520 basic_session_run_hooks.py:260] loss = 1.0562137, step = 234500 (3.155 sec)
I0804 21:43:04.863606 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5315
I0804 21:43:04.865169 140200711067520 basic_session_run_hooks.py:260] loss = 1.0662429, step = 234600 (3.172 sec)
I0804 21:43:08.058506 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2994
I0804 21:43:08.059821 140200711067520 basic_session_run_hooks.py:260] loss = 1.0629793, step = 234700 (3.195 sec)
I0804 21:43:11.222806 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6023
I0804 21:43:11.224182 140200711067520 basic_session_run_hooks.py:260] loss = 1.0674576, step = 234800 (3.164 sec)
I0804 21:43:14.418803 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2892
I0804 21:43:14.420108 140200711067520 basic_session_run_hooks.py:260] loss = 1.1022822, step = 234900 (3.196 sec)
I0804 21:43:17.610200 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 235000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:43:17.911176 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:43:17.956567 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2662
I0804 21:43:17.957699 140200711067520 basic_session_run_hooks.py:260] loss = 1.0416873, step = 235000 (3.538 sec)
I0804 21:43:21.182535 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9989
I0804 21:43:21.183922 140200711067520 basic_session_run_hooks.py:260] loss = 1.1116048, step = 235100 (3.226 sec)
I0804 21:43:24.375083 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3229
I0804 21:43:24.376516 140200711067520 basic_session_run_hooks.py:260] loss = 1.1308308, step = 235200 (3.193 sec)
I0804 21:43:27.584599 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1572
I0804 21:43:27.586100 140200711067520 basic_session_run_hooks.py:260] loss = 1.0439668, step = 235300 (3.210 sec)
I0804 21:43:30.796883 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1305
I0804 21:43:30.798309 140200711067520 basic_session_run_hooks.py:260] loss = 1.0549482, step = 235400 (3.212 sec)
I0804 21:43:34.000137 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2182
I0804 21:43:34.001680 140200711067520 basic_session_run_hooks.py:260] loss = 1.1123662, step = 235500 (3.203 sec)
I0804 21:43:37.201861 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2333
I0804 21:43:37.203063 140200711067520 basic_session_run_hooks.py:260] loss = 1.1693407, step = 235600 (3.201 sec)
I0804 21:43:40.392493 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3419
I0804 21:43:40.394011 140200711067520 basic_session_run_hooks.py:260] loss = 1.0816078, step = 235700 (3.191 sec)
I0804 21:43:43.616811 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0142
I0804 21:43:43.618099 140200711067520 basic_session_run_hooks.py:260] loss = 0.96943605, step = 235800 (3.224 sec)
I0804 21:43:46.852608 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9042
I0804 21:43:46.853706 140200711067520 basic_session_run_hooks.py:260] loss = 1.0872508, step = 235900 (3.236 sec)
I0804 21:43:50.049604 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 236000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:43:50.353207 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:43:50.391006 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2612
I0804 21:43:50.391939 140200711067520 basic_session_run_hooks.py:260] loss = 1.0891769, step = 236000 (3.538 sec)
I0804 21:43:53.617671 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.992
I0804 21:43:53.619023 140200711067520 basic_session_run_hooks.py:260] loss = 1.093087, step = 236100 (3.227 sec)
I0804 21:43:56.834353 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0879
I0804 21:43:56.835499 140200711067520 basic_session_run_hooks.py:260] loss = 1.0390338, step = 236200 (3.216 sec)
I0804 21:44:00.046151 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1353
I0804 21:44:00.047582 140200711067520 basic_session_run_hooks.py:260] loss = 1.0573373, step = 236300 (3.212 sec)
I0804 21:44:03.251084 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2019
I0804 21:44:03.252528 140200711067520 basic_session_run_hooks.py:260] loss = 1.1174649, step = 236400 (3.205 sec)
I0804 21:44:06.462062 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.143
I0804 21:44:06.463677 140200711067520 basic_session_run_hooks.py:260] loss = 1.1591996, step = 236500 (3.211 sec)
I0804 21:44:09.641982 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4473
I0804 21:44:09.643536 140200711067520 basic_session_run_hooks.py:260] loss = 1.0244304, step = 236600 (3.180 sec)
I0804 21:44:12.787028 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7961
I0804 21:44:12.788276 140200711067520 basic_session_run_hooks.py:260] loss = 1.0708632, step = 236700 (3.145 sec)
I0804 21:44:15.985611 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2639
I0804 21:44:15.986950 140200711067520 basic_session_run_hooks.py:260] loss = 1.0375067, step = 236800 (3.199 sec)
I0804 21:44:19.307167 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.1064
I0804 21:44:19.308480 140200711067520 basic_session_run_hooks.py:260] loss = 1.0930641, step = 236900 (3.322 sec)
I0804 21:44:22.513531 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 237000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:44:22.807470 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:44:22.848519 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.2378
I0804 21:44:22.849591 140200711067520 basic_session_run_hooks.py:260] loss = 1.0999188, step = 237000 (3.541 sec)
I0804 21:44:26.072208 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0204
I0804 21:44:26.073639 140200711067520 basic_session_run_hooks.py:260] loss = 1.0312728, step = 237100 (3.224 sec)
I0804 21:44:29.282079 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1539
I0804 21:44:29.283569 140200711067520 basic_session_run_hooks.py:260] loss = 1.0648234, step = 237200 (3.210 sec)
I0804 21:44:32.511911 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9616
I0804 21:44:32.513397 140200711067520 basic_session_run_hooks.py:260] loss = 1.117892, step = 237300 (3.230 sec)
I0804 21:44:35.748513 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8967
I0804 21:44:35.749749 140200711067520 basic_session_run_hooks.py:260] loss = 1.0438336, step = 237400 (3.236 sec)
I0804 21:44:38.976457 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9797
I0804 21:44:38.977872 140200711067520 basic_session_run_hooks.py:260] loss = 1.1113886, step = 237500 (3.228 sec)
I0804 21:44:42.197185 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0484
I0804 21:44:42.198271 140200711067520 basic_session_run_hooks.py:260] loss = 1.0123738, step = 237600 (3.220 sec)
I0804 21:44:45.439758 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8395
I0804 21:44:45.441124 140200711067520 basic_session_run_hooks.py:260] loss = 1.0482165, step = 237700 (3.243 sec)
I0804 21:44:48.652160 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1295
I0804 21:44:48.653520 140200711067520 basic_session_run_hooks.py:260] loss = 1.074094, step = 237800 (3.212 sec)
I0804 21:44:51.877257 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0068
I0804 21:44:51.878750 140200711067520 basic_session_run_hooks.py:260] loss = 1.0437338, step = 237900 (3.225 sec)
I0804 21:44:55.033199 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 238000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:44:55.334051 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:44:55.372462 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6106
I0804 21:44:55.373724 140200711067520 basic_session_run_hooks.py:260] loss = 1.0469205, step = 238000 (3.495 sec)
I0804 21:44:58.613392 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8554
I0804 21:44:58.614610 140200711067520 basic_session_run_hooks.py:260] loss = 1.0829542, step = 238100 (3.241 sec)
I0804 21:45:01.795826 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4224
I0804 21:45:01.797073 140200711067520 basic_session_run_hooks.py:260] loss = 1.0306995, step = 238200 (3.182 sec)
I0804 21:45:05.011282 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0998
I0804 21:45:05.012765 140200711067520 basic_session_run_hooks.py:260] loss = 1.1021671, step = 238300 (3.216 sec)
I0804 21:45:08.255289 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8261
I0804 21:45:08.257132 140200711067520 basic_session_run_hooks.py:260] loss = 1.0611577, step = 238400 (3.244 sec)
I0804 21:45:11.490611 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9087
I0804 21:45:11.491862 140200711067520 basic_session_run_hooks.py:260] loss = 1.0874064, step = 238500 (3.235 sec)
I0804 21:45:14.622144 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.9334
I0804 21:45:14.623574 140200711067520 basic_session_run_hooks.py:260] loss = 1.0911086, step = 238600 (3.132 sec)
I0804 21:45:17.760119 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.8676
I0804 21:45:17.761200 140200711067520 basic_session_run_hooks.py:260] loss = 1.0138427, step = 238700 (3.138 sec)
I0804 21:45:20.907254 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7749
I0804 21:45:20.908784 140200711067520 basic_session_run_hooks.py:260] loss = 1.0666616, step = 238800 (3.148 sec)
I0804 21:45:24.052583 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7934
I0804 21:45:24.053934 140200711067520 basic_session_run_hooks.py:260] loss = 1.0922306, step = 238900 (3.145 sec)
I0804 21:45:27.179445 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 239000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:45:27.483794 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:45:27.518613 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.851
I0804 21:45:27.519754 140200711067520 basic_session_run_hooks.py:260] loss = 1.0669379, step = 239000 (3.466 sec)
I0804 21:45:30.664836 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7844
I0804 21:45:30.665939 140200711067520 basic_session_run_hooks.py:260] loss = 1.1269906, step = 239100 (3.146 sec)
I0804 21:45:33.803495 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.861
I0804 21:45:33.804798 140200711067520 basic_session_run_hooks.py:260] loss = 1.0596148, step = 239200 (3.139 sec)
I0804 21:45:36.991399 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3683
I0804 21:45:36.992780 140200711067520 basic_session_run_hooks.py:260] loss = 1.0643686, step = 239300 (3.188 sec)
I0804 21:45:40.153000 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6298
I0804 21:45:40.154393 140200711067520 basic_session_run_hooks.py:260] loss = 0.9834525, step = 239400 (3.162 sec)
I0804 21:45:43.314146 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6339
I0804 21:45:43.315265 140200711067520 basic_session_run_hooks.py:260] loss = 1.0852195, step = 239500 (3.161 sec)
I0804 21:45:46.469586 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6916
I0804 21:45:46.470856 140200711067520 basic_session_run_hooks.py:260] loss = 1.0479075, step = 239600 (3.156 sec)
I0804 21:45:49.648491 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4573
I0804 21:45:49.649867 140200711067520 basic_session_run_hooks.py:260] loss = 1.0235535, step = 239700 (3.179 sec)
I0804 21:45:52.816260 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5677
I0804 21:45:52.817631 140200711067520 basic_session_run_hooks.py:260] loss = 1.0414634, step = 239800 (3.168 sec)
I0804 21:45:56.009608 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3152
I0804 21:45:56.010935 140200711067520 basic_session_run_hooks.py:260] loss = 1.136335, step = 239900 (3.193 sec)
I0804 21:45:59.173107 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 240000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:45:59.467769 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:45:59.508716 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5784
I0804 21:45:59.509741 140200711067520 basic_session_run_hooks.py:260] loss = 1.0130584, step = 240000 (3.499 sec)
I0804 21:46:02.705387 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2828
I0804 21:46:02.706692 140200711067520 basic_session_run_hooks.py:260] loss = 1.0709659, step = 240100 (3.197 sec)
I0804 21:46:05.880903 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.491
I0804 21:46:05.882242 140200711067520 basic_session_run_hooks.py:260] loss = 1.0985026, step = 240200 (3.176 sec)
I0804 21:46:09.056186 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4933
I0804 21:46:09.057739 140200711067520 basic_session_run_hooks.py:260] loss = 1.0706861, step = 240300 (3.175 sec)
I0804 21:46:12.253840 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2729
I0804 21:46:12.255214 140200711067520 basic_session_run_hooks.py:260] loss = 1.018328, step = 240400 (3.197 sec)
I0804 21:46:15.467080 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1213
I0804 21:46:15.468443 140200711067520 basic_session_run_hooks.py:260] loss = 1.0235646, step = 240500 (3.213 sec)
I0804 21:46:18.655800 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3605
I0804 21:46:18.657149 140200711067520 basic_session_run_hooks.py:260] loss = 1.100911, step = 240600 (3.189 sec)
I0804 21:46:21.865941 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1513
I0804 21:46:21.867340 140200711067520 basic_session_run_hooks.py:260] loss = 1.1124645, step = 240700 (3.210 sec)
I0804 21:46:25.060321 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.305
I0804 21:46:25.061862 140200711067520 basic_session_run_hooks.py:260] loss = 1.0312389, step = 240800 (3.195 sec)
I0804 21:46:28.278826 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0702
I0804 21:46:28.280752 140200711067520 basic_session_run_hooks.py:260] loss = 1.0192027, step = 240900 (3.219 sec)
I0804 21:46:31.406462 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 241000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:46:31.700452 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:46:31.738887 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.901
I0804 21:46:31.739915 140200711067520 basic_session_run_hooks.py:260] loss = 1.0117443, step = 241000 (3.459 sec)
I0804 21:46:34.917383 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4617
I0804 21:46:34.918823 140200711067520 basic_session_run_hooks.py:260] loss = 1.0754311, step = 241100 (3.179 sec)
I0804 21:46:38.077808 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6416
I0804 21:46:38.079228 140200711067520 basic_session_run_hooks.py:260] loss = 1.0486562, step = 241200 (3.160 sec)
I0804 21:46:41.237542 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6482
I0804 21:46:41.239070 140200711067520 basic_session_run_hooks.py:260] loss = 1.1193267, step = 241300 (3.160 sec)
I0804 21:46:44.412142 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4998
I0804 21:46:44.413800 140200711067520 basic_session_run_hooks.py:260] loss = 0.98590237, step = 241400 (3.175 sec)
I0804 21:46:47.558609 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.782
I0804 21:46:47.559746 140200711067520 basic_session_run_hooks.py:260] loss = 1.0820919, step = 241500 (3.146 sec)
I0804 21:46:50.745961 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3737
I0804 21:46:50.747091 140200711067520 basic_session_run_hooks.py:260] loss = 0.9946585, step = 241600 (3.187 sec)
I0804 21:46:53.987443 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8501
I0804 21:46:53.988624 140200711067520 basic_session_run_hooks.py:260] loss = 1.0404077, step = 241700 (3.242 sec)
I0804 21:46:57.173147 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3903
I0804 21:46:57.174671 140200711067520 basic_session_run_hooks.py:260] loss = 1.1369101, step = 241800 (3.186 sec)
I0804 21:47:00.359677 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3819
I0804 21:47:00.361214 140200711067520 basic_session_run_hooks.py:260] loss = 1.0505054, step = 241900 (3.187 sec)
I0804 21:47:03.507377 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 242000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:47:03.815570 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 21:47:03.816919 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 21:47:03.962552 140200711067520 estimator.py:1145] Calling model_fn.
I0804 21:47:03.963644 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 21:47:03.964045 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 21:47:03.964134 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 21:47:03.964214 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 21:47:03.964280 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 21:47:03.964360 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 21:47:03.964439 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 21:47:04.051969 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 21:47:04.107337 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 21:47:04.253585 140200711067520 t2t_model.py:2172] Building model body
I0804 21:47:04.938382 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 21:47:05.914096 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 21:47:05.932654 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T21:47:05Z
I0804 21:47:06.098258 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 21:47:06.098989: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:47:06.099388: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 21:47:06.099491: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 21:47:06.099515: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 21:47:06.099544: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 21:47:06.099564: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 21:47:06.099588: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 21:47:06.099610: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 21:47:06.099630: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 21:47:06.099735: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:47:06.100126: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:47:06.100451: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 21:47:06.100494: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 21:47:06.100508: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 21:47:06.100518: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 21:47:06.100802: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:47:06.101191: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:47:06.101534: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 21:47:06.102803 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-242000
I0804 21:47:06.327364 140200711067520 session_manager.py:500] Running local_init_op.
I0804 21:47:06.377162 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 21:47:12.427685 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 21:47:17.808200 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 21:47:23.208933 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 21:47:28.519458 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 21:47:33.827487 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 21:47:39.261294 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 21:47:44.644403 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 21:47:50.017316 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 21:47:55.360442 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 21:48:00.181250 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-21:48:00
I0804 21:48:00.181495 140200711067520 estimator.py:2039] Saving dict for global step 242000: global_step = 242000, loss = 1.158485, metrics-paper_generation_problem/targets/accuracy = 0.67867917, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.887068, metrics-paper_generation_problem/targets/approx_bleu_score = 0.49482146, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1585196, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.58801, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.69926846
I0804 21:48:00.181989 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 242000: experiment/transformer/transformer_small/output/model.ckpt-242000
I0804 21:48:00.237398 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 1.67007
I0804 21:48:00.238497 140200711067520 basic_session_run_hooks.py:260] loss = 1.0311191, step = 242000 (59.877 sec)
I0804 21:48:03.451106 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1173
I0804 21:48:03.452609 140200711067520 basic_session_run_hooks.py:260] loss = 1.0118092, step = 242100 (3.214 sec)
I0804 21:48:06.663685 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1277
I0804 21:48:06.665140 140200711067520 basic_session_run_hooks.py:260] loss = 0.97827464, step = 242200 (3.213 sec)
I0804 21:48:09.850941 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3748
I0804 21:48:09.852441 140200711067520 basic_session_run_hooks.py:260] loss = 1.1684557, step = 242300 (3.187 sec)
I0804 21:48:13.072041 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0452
I0804 21:48:13.073511 140200711067520 basic_session_run_hooks.py:260] loss = 1.1079253, step = 242400 (3.221 sec)
I0804 21:48:16.323067 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7596
I0804 21:48:16.324456 140200711067520 basic_session_run_hooks.py:260] loss = 1.0384694, step = 242500 (3.251 sec)
I0804 21:48:19.555159 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9396
I0804 21:48:19.556547 140200711067520 basic_session_run_hooks.py:260] loss = 1.0251623, step = 242600 (3.232 sec)
I0804 21:48:22.796639 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8504
I0804 21:48:22.797828 140200711067520 basic_session_run_hooks.py:260] loss = 1.1196532, step = 242700 (3.241 sec)
I0804 21:48:26.040591 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8267
I0804 21:48:26.041953 140200711067520 basic_session_run_hooks.py:260] loss = 0.9685662, step = 242800 (3.244 sec)
I0804 21:48:29.285764 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8147
I0804 21:48:29.287200 140200711067520 basic_session_run_hooks.py:260] loss = 1.0057174, step = 242900 (3.245 sec)
I0804 21:48:32.495259 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 243000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:48:32.802822 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:48:32.846061 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.0875
I0804 21:48:32.847136 140200711067520 basic_session_run_hooks.py:260] loss = 1.1071194, step = 243000 (3.560 sec)
I0804 21:48:36.058972 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1245
I0804 21:48:36.060117 140200711067520 basic_session_run_hooks.py:260] loss = 1.0428252, step = 243100 (3.213 sec)
I0804 21:48:39.263243 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2084
I0804 21:48:39.264497 140200711067520 basic_session_run_hooks.py:260] loss = 1.0617425, step = 243200 (3.204 sec)
I0804 21:48:42.498184 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9125
I0804 21:48:42.499596 140200711067520 basic_session_run_hooks.py:260] loss = 1.0753982, step = 243300 (3.235 sec)
I0804 21:48:45.677229 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4561
I0804 21:48:45.678302 140200711067520 basic_session_run_hooks.py:260] loss = 1.0918909, step = 243400 (3.179 sec)
I0804 21:48:48.868138 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3391
I0804 21:48:48.869381 140200711067520 basic_session_run_hooks.py:260] loss = 1.02013, step = 243500 (3.191 sec)
I0804 21:48:52.061055 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3193
I0804 21:48:52.062501 140200711067520 basic_session_run_hooks.py:260] loss = 1.0496832, step = 243600 (3.193 sec)
I0804 21:48:55.227055 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5855
I0804 21:48:55.228598 140200711067520 basic_session_run_hooks.py:260] loss = 1.0688226, step = 243700 (3.166 sec)
I0804 21:48:58.424380 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.276
I0804 21:48:58.425984 140200711067520 basic_session_run_hooks.py:260] loss = 1.0553972, step = 243800 (3.197 sec)
I0804 21:49:01.592632 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5632
I0804 21:49:01.594002 140200711067520 basic_session_run_hooks.py:260] loss = 1.0292231, step = 243900 (3.168 sec)
I0804 21:49:04.724610 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 244000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:49:05.034496 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:49:05.075029 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.7158
I0804 21:49:05.075978 140200711067520 basic_session_run_hooks.py:260] loss = 1.084882, step = 244000 (3.482 sec)
I0804 21:49:08.279037 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2111
I0804 21:49:08.280783 140200711067520 basic_session_run_hooks.py:260] loss = 1.0838699, step = 244100 (3.205 sec)
I0804 21:49:11.463529 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4026
I0804 21:49:11.465234 140200711067520 basic_session_run_hooks.py:260] loss = 1.0864109, step = 244200 (3.184 sec)
I0804 21:49:14.627151 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.609
I0804 21:49:14.628484 140200711067520 basic_session_run_hooks.py:260] loss = 1.1059449, step = 244300 (3.163 sec)
I0804 21:49:17.790354 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6136
I0804 21:49:17.791972 140200711067520 basic_session_run_hooks.py:260] loss = 0.9921102, step = 244400 (3.163 sec)
I0804 21:49:20.993198 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2221
I0804 21:49:20.994654 140200711067520 basic_session_run_hooks.py:260] loss = 1.1120347, step = 244500 (3.203 sec)
I0804 21:49:24.167629 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5017
I0804 21:49:24.169000 140200711067520 basic_session_run_hooks.py:260] loss = 1.1086354, step = 244600 (3.174 sec)
I0804 21:49:27.354388 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3798
I0804 21:49:27.355805 140200711067520 basic_session_run_hooks.py:260] loss = 1.0182055, step = 244700 (3.187 sec)
I0804 21:49:30.537302 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4178
I0804 21:49:30.538809 140200711067520 basic_session_run_hooks.py:260] loss = 1.0238302, step = 244800 (3.183 sec)
I0804 21:49:33.774222 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8936
I0804 21:49:33.775452 140200711067520 basic_session_run_hooks.py:260] loss = 1.1151732, step = 244900 (3.237 sec)
I0804 21:49:36.948932 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 245000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:49:37.251060 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:49:37.286807 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.4688
I0804 21:49:37.287880 140200711067520 basic_session_run_hooks.py:260] loss = 0.9793405, step = 245000 (3.512 sec)
I0804 21:49:40.514965 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9776
I0804 21:49:40.516059 140200711067520 basic_session_run_hooks.py:260] loss = 1.0599868, step = 245100 (3.228 sec)
I0804 21:49:43.708142 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3169
I0804 21:49:43.709439 140200711067520 basic_session_run_hooks.py:260] loss = 1.1070263, step = 245200 (3.193 sec)
I0804 21:49:46.914499 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1882
I0804 21:49:46.915846 140200711067520 basic_session_run_hooks.py:260] loss = 1.0083429, step = 245300 (3.206 sec)
I0804 21:49:50.128821 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1106
I0804 21:49:50.130345 140200711067520 basic_session_run_hooks.py:260] loss = 0.95696205, step = 245400 (3.214 sec)
I0804 21:49:53.351750 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0277
I0804 21:49:53.353066 140200711067520 basic_session_run_hooks.py:260] loss = 1.0594735, step = 245500 (3.223 sec)
I0804 21:49:56.560221 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1674
I0804 21:49:56.561326 140200711067520 basic_session_run_hooks.py:260] loss = 1.0052801, step = 245600 (3.208 sec)
I0804 21:49:59.795258 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9117
I0804 21:49:59.796679 140200711067520 basic_session_run_hooks.py:260] loss = 1.1649494, step = 245700 (3.235 sec)
I0804 21:50:02.988245 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3186
I0804 21:50:02.989299 140200711067520 basic_session_run_hooks.py:260] loss = 1.0641108, step = 245800 (3.193 sec)
I0804 21:50:06.152301 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.605
I0804 21:50:06.153884 140200711067520 basic_session_run_hooks.py:260] loss = 1.0329115, step = 245900 (3.165 sec)
I0804 21:50:09.308171 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 246000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:50:09.624747 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:50:09.660280 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.5062
I0804 21:50:09.661396 140200711067520 basic_session_run_hooks.py:260] loss = 1.0488344, step = 246000 (3.508 sec)
I0804 21:50:12.881713 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0423
I0804 21:50:12.883068 140200711067520 basic_session_run_hooks.py:260] loss = 1.1222789, step = 246100 (3.222 sec)
I0804 21:50:16.074593 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.32
I0804 21:50:16.075841 140200711067520 basic_session_run_hooks.py:260] loss = 1.0879499, step = 246200 (3.193 sec)
I0804 21:50:19.248188 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.5098
I0804 21:50:19.249301 140200711067520 basic_session_run_hooks.py:260] loss = 1.1017461, step = 246300 (3.173 sec)
I0804 21:50:22.412262 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.6048
I0804 21:50:22.413403 140200711067520 basic_session_run_hooks.py:260] loss = 1.0248648, step = 246400 (3.164 sec)
I0804 21:50:25.609134 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2805
I0804 21:50:25.610527 140200711067520 basic_session_run_hooks.py:260] loss = 1.0770847, step = 246500 (3.197 sec)
I0804 21:50:28.815371 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1892
I0804 21:50:28.816637 140200711067520 basic_session_run_hooks.py:260] loss = 1.0767981, step = 246600 (3.206 sec)
I0804 21:50:31.966548 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.7344
I0804 21:50:31.967887 140200711067520 basic_session_run_hooks.py:260] loss = 1.0686297, step = 246700 (3.151 sec)
I0804 21:50:35.265190 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.3154
I0804 21:50:35.266546 140200711067520 basic_session_run_hooks.py:260] loss = 1.0401849, step = 246800 (3.299 sec)
I0804 21:50:38.462337 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2778
I0804 21:50:38.463799 140200711067520 basic_session_run_hooks.py:260] loss = 1.032178, step = 246900 (3.197 sec)
I0804 21:50:41.652959 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 247000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:50:41.950251 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:50:41.990879 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.3403
I0804 21:50:41.991950 140200711067520 basic_session_run_hooks.py:260] loss = 1.0918233, step = 247000 (3.528 sec)
I0804 21:50:45.209505 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0695
I0804 21:50:45.210887 140200711067520 basic_session_run_hooks.py:260] loss = 1.0414871, step = 247100 (3.219 sec)
I0804 21:50:48.425673 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0927
I0804 21:50:48.427007 140200711067520 basic_session_run_hooks.py:260] loss = 1.0184749, step = 247200 (3.216 sec)
I0804 21:50:52.072402 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 27.4219
I0804 21:50:52.073872 140200711067520 basic_session_run_hooks.py:260] loss = 1.113687, step = 247300 (3.647 sec)
I0804 21:50:55.419208 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 29.8792
I0804 21:50:55.420756 140200711067520 basic_session_run_hooks.py:260] loss = 1.011707, step = 247400 (3.347 sec)
I0804 21:50:58.631895 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1266
I0804 21:50:58.633211 140200711067520 basic_session_run_hooks.py:260] loss = 1.0247351, step = 247500 (3.212 sec)
I0804 21:51:01.878511 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8014
I0804 21:51:01.879703 140200711067520 basic_session_run_hooks.py:260] loss = 1.036893, step = 247600 (3.246 sec)
I0804 21:51:05.108237 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9623
I0804 21:51:05.109721 140200711067520 basic_session_run_hooks.py:260] loss = 1.1814476, step = 247700 (3.230 sec)
I0804 21:51:08.335161 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9892
I0804 21:51:08.336711 140200711067520 basic_session_run_hooks.py:260] loss = 1.0198172, step = 247800 (3.227 sec)
I0804 21:51:11.569360 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9195
I0804 21:51:11.570667 140200711067520 basic_session_run_hooks.py:260] loss = 1.0243868, step = 247900 (3.234 sec)
I0804 21:51:14.782967 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 248000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:51:15.085118 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:51:15.127589 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.1038
I0804 21:51:15.128804 140200711067520 basic_session_run_hooks.py:260] loss = 1.0165476, step = 248000 (3.558 sec)
I0804 21:51:18.374958 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7944
I0804 21:51:18.376135 140200711067520 basic_session_run_hooks.py:260] loss = 1.0150435, step = 248100 (3.247 sec)
I0804 21:51:21.614459 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8693
I0804 21:51:21.615649 140200711067520 basic_session_run_hooks.py:260] loss = 1.1167299, step = 248200 (3.240 sec)
I0804 21:51:24.856394 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8455
I0804 21:51:24.857860 140200711067520 basic_session_run_hooks.py:260] loss = 1.0447816, step = 248300 (3.242 sec)
I0804 21:51:28.106035 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.7727
I0804 21:51:28.107668 140200711067520 basic_session_run_hooks.py:260] loss = 1.1144595, step = 248400 (3.250 sec)
I0804 21:51:31.346330 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.8612
I0804 21:51:31.347870 140200711067520 basic_session_run_hooks.py:260] loss = 1.1292459, step = 248500 (3.240 sec)
I0804 21:51:34.557864 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.1377
I0804 21:51:34.559124 140200711067520 basic_session_run_hooks.py:260] loss = 1.0045581, step = 248600 (3.211 sec)
I0804 21:51:37.792023 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.92
I0804 21:51:37.793512 140200711067520 basic_session_run_hooks.py:260] loss = 1.0909798, step = 248700 (3.234 sec)
I0804 21:51:41.013727 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0395
I0804 21:51:41.014953 140200711067520 basic_session_run_hooks.py:260] loss = 0.9874409, step = 248800 (3.221 sec)
I0804 21:51:44.237439 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0203
I0804 21:51:44.238864 140200711067520 basic_session_run_hooks.py:260] loss = 1.0409068, step = 248900 (3.224 sec)
I0804 21:51:47.398524 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 249000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:51:47.697914 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:51:47.733337 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 28.6046
I0804 21:51:47.734453 140200711067520 basic_session_run_hooks.py:260] loss = 1.2293057, step = 249000 (3.496 sec)
I0804 21:51:50.958437 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.0071
I0804 21:51:50.959923 140200711067520 basic_session_run_hooks.py:260] loss = 1.0172132, step = 249100 (3.225 sec)
I0804 21:51:54.186296 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 30.9801
I0804 21:51:54.187634 140200711067520 basic_session_run_hooks.py:260] loss = 1.0418226, step = 249200 (3.228 sec)
I0804 21:51:57.362849 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4806
I0804 21:51:57.364353 140200711067520 basic_session_run_hooks.py:260] loss = 1.0257041, step = 249300 (3.177 sec)
I0804 21:52:00.555961 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.3174
I0804 21:52:00.557339 140200711067520 basic_session_run_hooks.py:260] loss = 0.9991082, step = 249400 (3.193 sec)
I0804 21:52:03.736152 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.4448
I0804 21:52:03.737734 140200711067520 basic_session_run_hooks.py:260] loss = 1.0921255, step = 249500 (3.180 sec)
I0804 21:52:06.908018 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.527
I0804 21:52:06.909434 140200711067520 basic_session_run_hooks.py:260] loss = 1.0971957, step = 249600 (3.172 sec)
I0804 21:52:10.099232 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.336
I0804 21:52:10.100273 140200711067520 basic_session_run_hooks.py:260] loss = 1.0196719, step = 249700 (3.191 sec)
I0804 21:52:13.303578 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2078
I0804 21:52:13.305107 140200711067520 basic_session_run_hooks.py:260] loss = 1.042994, step = 249800 (3.205 sec)
I0804 21:52:16.501100 140200711067520 basic_session_run_hooks.py:692] global_step/sec: 31.2743
I0804 21:52:16.502824 140200711067520 basic_session_run_hooks.py:260] loss = 1.0992345, step = 249900 (3.198 sec)
I0804 21:52:19.720271 140200711067520 basic_session_run_hooks.py:606] Saving checkpoints for 250000 into experiment/transformer/transformer_small/output/model.ckpt.
I0804 21:52:20.020960 140200711067520 training.py:527] Skip the current checkpoint eval due to throttle secs (600 secs).
I0804 21:52:20.029833 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 21:52:20.031070 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 21:52:20.175315 140200711067520 estimator.py:1145] Calling model_fn.
I0804 21:52:20.176227 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 21:52:20.176643 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 21:52:20.176737 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 21:52:20.176818 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 21:52:20.176885 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 21:52:20.176966 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 21:52:20.177031 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 21:52:20.262790 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 21:52:20.321675 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 21:52:20.724619 140200711067520 t2t_model.py:2172] Building model body
I0804 21:52:21.403978 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 21:52:22.101516 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 21:52:22.120038 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T21:52:22Z
I0804 21:52:22.489091 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 21:52:22.489784: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:52:22.490255: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 21:52:22.490344: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 21:52:22.490372: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 21:52:22.490393: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 21:52:22.490439: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 21:52:22.490462: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 21:52:22.490481: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 21:52:22.490501: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 21:52:22.490613: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:52:22.491005: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:52:22.491318: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 21:52:22.491360: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 21:52:22.491374: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 21:52:22.491384: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 21:52:22.491682: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:52:22.492065: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:52:22.492391: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 21:52:22.494179 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-250000
I0804 21:52:22.699904 140200711067520 session_manager.py:500] Running local_init_op.
I0804 21:52:22.753980 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 21:52:28.791632 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 21:52:34.121817 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 21:52:39.491192 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 21:52:44.877773 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 21:52:50.320030 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 21:52:55.758518 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 21:53:01.118310 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 21:53:06.478620 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 21:53:11.816269 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 21:53:16.664833 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-21:53:16
I0804 21:53:16.665058 140200711067520 estimator.py:2039] Saving dict for global step 250000: global_step = 250000, loss = 1.1575909, metrics-paper_generation_problem/targets/accuracy = 0.67876077, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8869048, metrics-paper_generation_problem/targets/approx_bleu_score = 0.49531543, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1576262, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5884521, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.6995779
I0804 21:53:16.665603 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 250000: experiment/transformer/transformer_small/output/model.ckpt-250000
I0804 21:53:16.910505 140200711067520 estimator.py:368] Loss for final step: 1.0043204.
I0804 21:53:16.918406 140200711067520 problem.py:644] Reading data files from experiment/transformer/transformer_small/data/paper_generation_problem-dev*
I0804 21:53:16.919665 140200711067520 problem.py:670] partition: 0 num_data_files: 1
I0804 21:53:17.068397 140200711067520 estimator.py:1145] Calling model_fn.
I0804 21:53:17.069321 140200711067520 t2t_model.py:2172] Setting T2TModel mode to 'eval'
I0804 21:53:17.069730 140200711067520 t2t_model.py:2172] Setting hparams.dropout to 0.0
I0804 21:53:17.069826 140200711067520 t2t_model.py:2172] Setting hparams.label_smoothing to 0.0
I0804 21:53:17.069909 140200711067520 t2t_model.py:2172] Setting hparams.layer_prepostprocess_dropout to 0.0
I0804 21:53:17.069978 140200711067520 t2t_model.py:2172] Setting hparams.symbol_dropout to 0.0
I0804 21:53:17.070062 140200711067520 t2t_model.py:2172] Setting hparams.attention_dropout to 0.0
I0804 21:53:17.070136 140200711067520 t2t_model.py:2172] Setting hparams.relu_dropout to 0.0
I0804 21:53:17.153735 140200711067520 api.py:255] Using variable initializer: uniform_unit_scaling
I0804 21:53:17.210179 140200711067520 t2t_model.py:2172] Transforming feature 'targets' with symbol_modality_258_256.targets_bottom
I0804 21:53:17.343150 140200711067520 t2t_model.py:2172] Building model body
I0804 21:53:17.997269 140200711067520 t2t_model.py:2172] Transforming body output with symbol_modality_258_256.top
I0804 21:53:18.672821 140200711067520 estimator.py:1147] Done calling model_fn.
I0804 21:53:18.689967 140200711067520 evaluation.py:255] Starting evaluation at 2019-08-04T21:53:18Z
I0804 21:53:19.203626 140200711067520 monitored_session.py:240] Graph was finalized.
2019-08-04 21:53:19.204280: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:53:19.204704: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties: 
name: Tesla T4 major: 7 minor: 5 memoryClockRate(GHz): 1.59
pciBusID: 0000:00:04.0
2019-08-04 21:53:19.204789: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.0
2019-08-04 21:53:19.204813: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0
2019-08-04 21:53:19.204833: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10.0
2019-08-04 21:53:19.204852: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10.0
2019-08-04 21:53:19.204872: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10.0
2019-08-04 21:53:19.204891: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10.0
2019-08-04 21:53:19.204911: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
2019-08-04 21:53:19.205012: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:53:19.205396: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:53:19.205730: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-08-04 21:53:19.205773: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-04 21:53:19.205787: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187]      0 
2019-08-04 21:53:19.205797: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0:   N 
2019-08-04 21:53:19.206082: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:53:19.206501: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:1005] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-08-04 21:53:19.206835: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 14325 MB memory) -> physical GPU (device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5)
I0804 21:53:19.207969 140200711067520 saver.py:1280] Restoring parameters from experiment/transformer/transformer_small/output/model.ckpt-250000
I0804 21:53:19.389290 140200711067520 session_manager.py:500] Running local_init_op.
I0804 21:53:19.429021 140200711067520 session_manager.py:502] Done running local_init_op.
I0804 21:53:25.410798 140200711067520 evaluation.py:167] Evaluation [10/100]
I0804 21:53:30.759132 140200711067520 evaluation.py:167] Evaluation [20/100]
I0804 21:53:36.073087 140200711067520 evaluation.py:167] Evaluation [30/100]
I0804 21:53:41.378449 140200711067520 evaluation.py:167] Evaluation [40/100]
I0804 21:53:46.707476 140200711067520 evaluation.py:167] Evaluation [50/100]
I0804 21:53:52.034182 140200711067520 evaluation.py:167] Evaluation [60/100]
I0804 21:53:57.369941 140200711067520 evaluation.py:167] Evaluation [70/100]
I0804 21:54:02.755774 140200711067520 evaluation.py:167] Evaluation [80/100]
I0804 21:54:08.109610 140200711067520 evaluation.py:167] Evaluation [90/100]
I0804 21:54:12.930298 140200711067520 evaluation.py:275] Finished evaluation at 2019-08-04-21:54:12
I0804 21:54:12.930542 140200711067520 estimator.py:2039] Saving dict for global step 250000: global_step = 250000, loss = 1.1575909, metrics-paper_generation_problem/targets/accuracy = 0.67876077, metrics-paper_generation_problem/targets/accuracy_per_sequence = 0.0, metrics-paper_generation_problem/targets/accuracy_top5 = 0.8869048, metrics-paper_generation_problem/targets/approx_bleu_score = 0.49531543, metrics-paper_generation_problem/targets/neg_log_perplexity = -1.1576262, metrics-paper_generation_problem/targets/rouge_2_fscore = 0.5884521, metrics-paper_generation_problem/targets/rouge_L_fscore = 0.6995779
I0804 21:54:13.235319 140200711067520 estimator.py:2099] Saving 'checkpoint_path' summary for global step 250000: experiment/transformer/transformer_small/output/model.ckpt-250000
--> Train Completed. Files saved at experiment/transformer/transformer_small