Update scripts

scripts/new_prepare_pickled_dataset.py

@@ -0,0 +1,52 @@
+from portfolio_management.data.manager import Manager
+from portfolio_management.data.preprocessing import PCAPreprocessing
+from portfolio_management.data.dataset import DatasetManager
+
+
+def create_database(database_name, symbol_list, interval, start, end):
+    manager = Manager(
+        database_name=database_name,
+        reset_tables=True,
+    )
+    manager.insert(
+        symbol_list=symbol_list,
+        interval=interval,
+        start=start,
+        end=end,
+    )
+
+
+if __name__ == '__main__':
+
+    download_train_data = False
+    download_test_data = False
+
+    symbol_list = [
+        'USDCUSDT',
+        'BTCUSDT',
+        'ETHUSDT',
+        'BNBUSDT',
+        'LTCUSDT',
+        'XRPUSDT'
+    ]
+    interval = '1h'
+
+    train_database_name = 'USDT_train'
+    start = "2020-01-01"
+    end = "2020-12-31"
+    if download_train_data:
+        create_database(train_database_name, symbol_list, interval, start, end)
+
+    test_database_name = 'USDT_test'
+    start = "2021-01-01"
+    end = "2021-05-01"
+    if download_test_data:
+        create_database(test_database_name, symbol_list, interval, start, end)
+
+    preprocessing_class = PCAPreprocessing()
+
+    dataset_manager = DatasetManager(
+        train_database_name=train_database_name,
+        test_database_name=test_database_name,
+    )
+    dataset_manager.run(preprocessing_class)

scripts/prepare_pickled_dataset.py

@@ -5,9 +5,34 @@ from portfolio_management.data.retrieve import get_dataset
 from portfolio_management.data.preprocessing import get_pca_preprocessing_function
 
 
-if __name__ == '__main__':
+def main(database_name, symbol_list, interval, start, end):
+    manager = Manager(
+        database_name=database_name,
+        reset_tables=True,
+    )
 
-    database_name = 'USDT_train'
+
+    manager.insert(
+        symbol_list=symbol_list,
+        interval=interval,
+        start=start,
+        end=end,
+    )
+
+    preprocessing_function = get_pca_preprocessing_function()
+
+    dataset = get_dataset(  # todo find a way to prepare the test dataset the same way as the train set
+        database_name=database_name,
+        interval=interval,
+        preprocessing=preprocessing_function
+    )
+
+    datasets_folder_path = p.datasets_folder_path
+    path_dataset = datasets_folder_path.joinpath(database_name).with_suffix('.pkl')
+    pickle_dump(path_dataset, dataset)
+
+
+if __name__ == '__main__':
 
     symbol_list = [
         'USDCUSDT',
@@ -18,28 +43,13 @@ if __name__ == '__main__':
         'XRPUSDT'
     ]
     interval = '1h'
+
+    database_name = 'USDT_train'
     start = "2020-01-01"
     end = "2020-12-31"
+    # main(database_name, symbol_list, interval, start, end)
 
-    manager = Manager(
-        database_name=database_name,
-        reset_tables=True,
-    )
-    manager.insert(
-        symbol_list=symbol_list,
-        interval=interval,
-        start=start,
-        end=end,
-    )
-
-    preprocessing_function = get_pca_preprocessing_function()
-
-    dataset = get_dataset( # todo find a way to prepare the test dataset the same way as the train set
-        database_name=database_name,
-        interval=interval,
-        preprocessing=preprocessing_function
-    )
-
-    datasets_folder_path = p.datasets_folder_path
-    path_dataset = datasets_folder_path.joinpath(database_name).with_suffix('.pkl')
-    pickle_dump(path_dataset, dataset)
+    database_name = 'USDT_test'
+    start = "2021-01-01"
+    end = "2021-05-01"
+    main(database_name, symbol_list, interval, start, end)

scripts/run_sac.py

@@ -14,17 +14,17 @@ if __name__ == '__main__':  # can't test sac yet, there is 'inf' value in datase
     # first run 'prepare_pickled_dataset' !
     # todo find why it is slow
 
-    train = True
+    train = False
     dataset_name = 'USDT_train'
 
     env_kwargs = {
         'dataset_name': dataset_name,
         'num_steps': 100,
-        'fees': 0.002,
+        'fees': 0.005,
         'seed': 1,
         'step_size': 1,
         'chronologically': False,
-        'observation_size': 12,
+        'observation_size': 50,
         'stake_range': [100, 100],
     }
 
@@ -58,14 +58,17 @@ if __name__ == '__main__':  # can't test sac yet, there is 'inf' value in datase
         )
 
     else:
-        env_kwarg_test = {  # todo give the possibility to change the interval
+        dataset_name = 'USDT_test'
+        env_kwarg_test = {
+            'dataset_name': dataset_name,
+            'num_steps': None,
         }
         env_kwargs.update(**env_kwarg_test)
-        run_name = '20210608_204057_Portfolio-v0'
-        soft_actor_critic.evaluate(
+        run_name = '20210720_192852_Portfolio-v0'
+        soft_actor_critic.render(
             'Portfolio-v0',
             run_name=run_name,
             env_kwargs=env_kwargs,
-            num_episodes=100
+            num_episodes=1
         )
 

tests/soft_actor_critic/test_train_sac.py

@@ -5,7 +5,7 @@ from portfolio_management.io_utilities import pickle_dump
 from portfolio_management.soft_actor_critic import train
 
 
-def test_episode(download_and_pickle: bool = True):
+def test_episode(download_and_pickle: bool = False):
     database_name = 'test_episode'
 
     if download_and_pickle: