entailment works; add peek-dataset command

textattack/commands/attack/attack_args_helpers.py

@@ -333,6 +333,8 @@ def parse_dataset_from_args(args):
         dataset_args = args.dataset_from_nlp
         if ":" in dataset_args:
             dataset_args = dataset_args.split(":")
+        else:
+            dataset_args = (dataset_args,)
         dataset = textattack.datasets.HuggingFaceNLPDataset(
             *dataset_args, shuffle=args.shuffle
         )

textattack/commands/eval_model/__init__.py

@@ -0,0 +1 @@
+from .eval_model_command import EvalModelCommand

textattack/commands/peek_dataset.py

@@ -0,0 +1,73 @@
+from argparse import ArgumentDefaultsHelpFormatter, ArgumentParser
+
+import textattack
+import numpy as np
+import re
+
+from textattack.commands import TextAttackCommand
+from textattack.commands.attack.attack_args_helpers import add_dataset_args, parse_dataset_from_args
+
+
+def _cb(s):
+    return textattack.shared.utils.color_text(str(s), color="red", method="ansi")
+
+logger = textattack.shared.logger
+
+import re
+
+class PeekDatasetCommand(TextAttackCommand):
+    """
+    The peek dataset module:
+    
+        Takes a peek into a dataset in textattack.
+    """
+    def run(self, args):
+        UPPERCASE_LETTERS_REGEX = re.compile('[A-Z]')
+        
+        args.model = None # set model to None for parse_dataset_from_args to work
+        dataset = parse_dataset_from_args(args)
+        
+        num_words = []
+        attacked_texts = []
+        data_all_lowercased = True
+        outputs = []
+        for inputs, output in dataset:
+            at = textattack.shared.AttackedText(inputs)
+            if data_all_lowercased:
+                # Test if any of the letters in the string are lowercase.
+                if re.search(UPPERCASE_LETTERS_REGEX, at.text):
+                    data_all_lowercased = False
+            attacked_texts.append(at)
+            num_words.append(len(at.words))
+            outputs.append(output)
+        
+        logger.info(f'Number of samples: {_cb(len(attacked_texts))}')
+        logger.info(f'Number of words per input:')
+        num_words = np.array(num_words)
+        logger.info(f'\t{("total:").ljust(8)} {_cb(num_words.sum())}')
+        mean_words = f'{num_words.mean():.2f}'
+        logger.info(f'\t{("mean:").ljust(8)} {_cb(mean_words)}')
+        std_words = f'{num_words.std():.2f}'
+        logger.info(f'\t{("std:").ljust(8)} {_cb(std_words)}')
+        logger.info(f'\t{("min:").ljust(8)} {_cb(num_words.min())}')
+        logger.info(f'\t{("max:").ljust(8)} {_cb(num_words.max())}')
+        logger.info(f'Dataset lowercased: {_cb(data_all_lowercased)}')
+        
+        logger.info('First sample:')
+        print(attacked_texts[0].printable_text(), '\n')
+        logger.info('Last sample:')
+        print(attacked_texts[-1].printable_text(), '\n')
+        
+        outputs = set(outputs)
+        logger.info(f'Found {len(outputs)} distinct outputs:')
+        print(sorted(outputs))
+        
+
+    @staticmethod
+    def register_subcommand(main_parser: ArgumentParser):
+        parser = main_parser.add_parser(
+            "peek-dataset", help="show main statistics about a dataset",
+            formatter_class=ArgumentDefaultsHelpFormatter,
+        )
+        add_dataset_args(parser)
+        parser.set_defaults(func=PeekDatasetCommand())

textattack/commands/textattack_cli.py

@@ -8,9 +8,9 @@ from textattack.commands.augment import AugmentCommand
 from textattack.commands.benchmark_recipe import BenchmarkRecipeCommand
 from textattack.commands.eval_model import EvalModelCommand
 from textattack.commands.list_things import ListThingsCommand
+from textattack.commands.peek_dataset import PeekDatasetCommand
 from textattack.commands.train_model import TrainModelCommand
 
-
 def main():
     parser = argparse.ArgumentParser(
         "TextAttack CLI",
@@ -27,6 +27,7 @@ def main():
     EvalModelCommand.register_subcommand(subparsers)
     ListThingsCommand.register_subcommand(subparsers)
     TrainModelCommand.register_subcommand(subparsers)
+    PeekDatasetCommand.register_subcommand(subparsers)
 
     # Let's go
     args = parser.parse_args()

textattack/commands/train_model/run_training.py

@@ -7,6 +7,7 @@ from torch.utils.data import DataLoader, Dataset, RandomSampler
 import tqdm
 
 import textattack
+import transformers
 
 from .train_args_helpers import dataset_from_args, model_from_args
 
@@ -19,16 +20,15 @@ def make_directories(output_dir):
         os.makedirs(output_dir)
 
 
-def encode_batch(tokenizer, text_list):
-    try:
-        return tokenizer.encode_batch(text_list)
-    except AttributeError:
-        return [tokenizer.encode(text) for text in text_list]
+def batch_encode(tokenizer, text_list):
+    return [tokenizer.encode(text_input) for text_input in text_list]
+    # try:
+        # return tokenizer.batch_encode(text_list) # TODO get batch encoding to work with fast tokenizer
+    # except AttributeError:
+        # return [tokenizer.encode(text_input) for text_input in text_list]
 
 
 def train_model(args):
-    from transformers.optimization import AdamW, get_linear_schedule_with_warmup
-
     start_time = time.time()
     make_directories(args.output_dir)
 
@@ -38,8 +38,12 @@ def train_model(args):
     from tensorboardX import SummaryWriter
 
     tb_writer = SummaryWriter(args.output_dir)
-    args_dict = vars(args)
-    del args_dict["func"]
+    def is_writable_type(obj):
+        for ok_type in [bool, int, str, float]:
+            if isinstance(obj, ok_type): return True
+        return False
+    args_dict = {k: v for k,v in vars(args).items() if is_writable_type(v)}
+    
     tb_writer.add_hparams(args_dict, {})
 
     # Use Weights & Biases, if enabled.
@@ -88,9 +92,9 @@ def train_model(args):
     model = model_from_args(args, num_labels)
 
     logger.info(f"Tokenizing training data. (len: {train_examples_len})")
-    train_text_ids = encode_batch(model.tokenizer, train_text)
-    logger.info(f"Tokenizing test data (len: {len(eval_labels)})")
-    eval_text_ids = encode_batch(model.tokenizer, eval_text)
+    train_text_ids = batch_encode(model.tokenizer, train_text)
+    logger.info(f"Tokenizing eval data (len: {len(eval_labels)})")
+    eval_text_ids = batch_encode(model.tokenizer, eval_text)
     load_time = time.time()
     logger.info(f"Loaded data and tokenized in {load_time-start_time}s")
     # print(model)
@@ -122,9 +126,9 @@ def train_model(args):
         },
     ]
 
-    optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate)
+    optimizer = transformers.optimization.AdamW(optimizer_grouped_parameters, lr=args.learning_rate)
 
-    scheduler = get_linear_schedule_with_warmup(
+    scheduler = transformers.optimization.get_linear_schedule_with_warmup(
         optimizer,
         num_warmup_steps=args.warmup_proportion,
         num_training_steps=num_train_optimization_steps,
@@ -133,12 +137,12 @@ def train_model(args):
     global_step = 0
 
     logger.info("***** Running training *****")
-    logger.info("  Num examples = %d", train_examples_len)
-    logger.info("  Batch size = %d", args.batch_size)
-    logger.info("  Max sequence length = %d", args.max_length)
-    logger.info("  Num steps = %d", num_train_optimization_steps)
-    logger.info("  Num epochs = %d", args.num_train_epochs)
-    logger.info("  Learning rate = %d", args.learning_rate)
+    logger.info(f"\tNum examples = {train_examples_len}")
+    logger.info(f"\tBatch size = {args.batch_size}")
+    logger.info(f"\tMax sequence length = {args.max_length}")
+    logger.info(f"\tNum steps = {num_train_optimization_steps}")
+    logger.info(f"\tNum epochs = {args.num_train_epochs}")
+    logger.info(f"\tLearning rate = {args.learning_rate}")
 
     train_input_ids = np.array(train_text_ids)
     train_labels = np.array(train_labels)
@@ -219,6 +223,7 @@ def train_model(args):
         if args.grad_accum_steps > 1:
             loss = loss / args.grad_accum_steps
         loss.backward()
+        return loss
 
     for _ in tqdm.trange(int(args.num_train_epochs), desc="Epoch"):
         prog_bar = tqdm.tqdm(train_dataloader, desc="Iteration")
@@ -231,20 +236,20 @@ def train_model(args):
                 input_ids = {
                     k: torch.stack(v).T.to(device) for k, v in input_ids.items()
                 }
-            logits = textattack.shared.utils.model_predict(model, input_ids)
-            loss_fct = torch.nn.CrossEntropyLoss()
-            loss = torch.nn.CrossEntropyLoss()(logits, labels)
+            input_ids['labels'] = labels # @TODO change this back to calculate loss in this body
+            loss = model(**input_ids)[0] 
+            loss = loss_backward(loss)
+            
             if global_step % args.tb_writer_step == 0:
-                tb_writer.add_scalar("loss", loss, global_step)
-                tb_writer.add_scalar("lr", loss, global_step)
-            loss_backward(loss)
+                tb_writer.add_scalar("loss", loss.item(), global_step)
+                tb_writer.add_scalar("lr", scheduler.get_last_lr()[0], global_step)
             prog_bar.set_description(f"Loss {loss.item()}")
             if (step + 1) % args.grad_accum_steps == 0:
                 optimizer.step()
                 scheduler.step()
                 optimizer.zero_grad()
             # Save model checkpoint to file.
-            if global_step > 0 and global_step % args.checkpoint_steps == 0:
+            if global_step > 0 and (args.checkpoint_steps > 0) and (global_step % args.checkpoint_steps) == 0:
                 save_model_checkpoint()
 
             model.zero_grad()
@@ -273,3 +278,10 @@ def train_model(args):
                     f"Stopping early since it's been {args.early_stopping_epochs} steps since validation acc increased"
                 )
                 break
+    
+    # end of training, save tokenizer
+    try:
+        tokenizer.save_pretrained(args.output_dir)
+        logger.info(f"Saved tokenizer {tokenizer} to {args.output_dir}.")
+    except AttributeError:
+        logger.warn(f"Error: could not save tokenizer {tokenizer} to {args.output_dir}.")

textattack/commands/train_model/train_args_helpers.py

@@ -27,6 +27,11 @@ def dataset_from_args(args):
     """
     dataset_args = args.dataset.split(":")
     # TODO `HuggingFaceNLPDataset` -> `HuggingFaceDataset`
+    if args.dataset_train_split:
+            train_dataset = textattack.datasets.HuggingFaceNLPDataset(
+                *dataset_args, split=args.dataset_train_split
+            )
+    else:
         try:
             train_dataset = textattack.datasets.HuggingFaceNLPDataset(
                 *dataset_args, split="train"
@@ -35,9 +40,9 @@ def dataset_from_args(args):
             raise KeyError(f"Error: no `train` split found in `{args.dataset}` dataset")
     train_text, train_labels = prepare_dataset_for_training(train_dataset)
     
-    if args.dataset_split:
+    if args.dataset_dev_split:
         eval_dataset = textattack.datasets.HuggingFaceNLPDataset(
-            *dataset_args, split=args.dataset_split
+            *dataset_args, split=args.dataset_dev_split
         )
     else:
         # try common dev split names
@@ -61,7 +66,6 @@ def dataset_from_args(args):
                     )
     eval_text, eval_labels = prepare_dataset_for_training(eval_dataset)
 
-
     return train_text, train_labels, eval_text, eval_labels
 
 
@@ -94,7 +98,7 @@ def model_from_args(args, num_labels):
             config=config,
         )
         tokenizer = textattack.models.tokenizers.AutoTokenizer(
-            args.model, use_fast=False, max_length=args.max_length
+            args.model, use_fast=True, max_length=args.max_length
         )
         setattr(model, "tokenizer", tokenizer)
 

textattack/commands/train_model/train_model_command.py

@@ -49,10 +49,17 @@ class TrainModelCommand(TextAttackCommand):
             " ex: `glue:sst2` or `rotten_tomatoes`",
         )
         parser.add_argument(
-            "--dataset-split",
+            "--dataset-train-split", '--train-split',
             type=str,
             default='',
-            help="dataset split, if non-standard "
+            help="train dataset split, if non-standard "
+            "(can automatically detect 'train'",
+        )
+        parser.add_argument(
+            "--dataset-dev-split", '--dataset-val-split', '--dev-split',
+            type=str,
+            default='',
+            help="val dataset split, if non-standard "
             "(can automatically detect 'dev', 'validation', 'eval')",
         )
         parser.add_argument(
@@ -64,10 +71,9 @@ class TrainModelCommand(TextAttackCommand):
         parser.add_argument(
             "--checkpoint-steps",
             type=int,
-            default=5000,
-            help="save model after this many steps",
+            default=-1,
+            help="save model after this many steps (-1 for no checkpointing)",
         )
-
         parser.add_argument(
             "--checkpoint-every_epoch",
             action="store_true",
@@ -102,7 +108,6 @@ class TrainModelCommand(TextAttackCommand):
             help="Maximum length of a sequence (anything beyond this will "
             "be truncated)",
         )
-
         parser.add_argument(
             "--learning-rate",
             "--lr",
@@ -110,7 +115,6 @@ class TrainModelCommand(TextAttackCommand):
             default=2e-5,
             help="Learning rate for Adam Optimization",
         )
-
         parser.add_argument(
             "--grad-accum-steps",
             type=int,
@@ -118,33 +122,28 @@ class TrainModelCommand(TextAttackCommand):
             help="Number of steps to accumulate gradients before optimizing, "
             "advancing scheduler, etc.",
         )
-
         parser.add_argument(
             "--warmup-proportion",
             type=float,
             default=0.1,
             help="Warmup proportion for linear scheduling",
         )
-
         parser.add_argument(
             "--config-name",
             type=str,
             default="config.json",
             help="Filename to save BERT config as",
         )
-
         parser.add_argument(
             "--weights-name",
             type=str,
             default="pytorch_model.bin",
             help="Filename to save model weights as",
         )
-
         parser.add_argument(
             "--enable-wandb",
             default=False,
             action="store_true",
             help="log metrics to Weights & Biases",
         )
-
         parser.set_defaults(func=TrainModelCommand())

textattack/datasets/huggingface_nlp_dataset.py

@@ -2,6 +2,7 @@ import collections
 import random
 
 import nlp
+import textattack
 
 from textattack.datasets import TextAttackDataset
 from textattack.shared import AttackedText
@@ -67,6 +68,9 @@ class HuggingFaceNLPDataset(TextAttackDataset):
         dataset_columns=None,
         shuffle=False,
     ):
+        
+        subset_print_str = f", subset `{subset}`" if subset else ""
+        textattack.shared.logger.info(f"Loading `nlp` dataset `{name}`{subset_print_str}, split `{split}`.")
         dataset = nlp.load_dataset(name, subset)
         (
             self.input_columns,

textattack/models/tokenizers/auto_tokenizer.py

@@ -28,6 +28,8 @@ class AutoTokenizer(Tokenizer):
             name, use_fast=use_fast
         )
         self.max_length = max_length
+        print(f'AutoTokenizer using {"fast" if use_fast else "slow"} tokenizer')
+        self.save_pretrained = self.tokenizer.save_pretrained
 
     def encode(self, input_text):
         """ Encodes ``input_text``.
@@ -36,19 +38,24 @@ class AutoTokenizer(Tokenizer):
         model takes 1 or multiple inputs. The ``transformers.AutoTokenizer``
         will automatically handle either case.
         """
+        if isinstance(input_text, str):
+            input_text = (input_text, )
         encoded_text = self.tokenizer.encode_plus(
-            input_text,
+            *input_text,
             max_length=self.max_length,
             add_special_tokens=True,
             pad_to_max_length=True,
+            truncation=True
         )
         return dict(encoded_text)
 
-    def encode_batch(self, input_text_list):
+    def batch_encode(self, input_text_list):
         """ The batch equivalent of ``encode``."""
-        if hasattr(self.tokenizer, "encode_batch"):
-            return self.tokenizer.encode_batch(
+        if hasattr(self.tokenizer, "batch_encode_plus"):
+            print('utilizing batch encode')
+            return self.tokenizer.batch_encode_plus(
                 input_text_list,
+                truncation=True,
                 max_length=self.max_length,
                 add_special_tokens=True,
                 pad_to_max_length=True,

textattack/models/tokenizers/spacy_tokenizer.py

@@ -21,6 +21,8 @@ class SpacyTokenizer(Tokenizer):
 
     def convert_text_to_tokens(self, text):
         if isinstance(text, tuple):
+            if len(text) > 1:
+                raise TypeError('Cannot train LSTM/CNN models with multi-sequence inputs.')
             text = text[0]
         if not isinstance(text, str):
             raise TypeError(