remove all the print() statements and use some form of logging instead

plot.py

@@ -1,3 +1,14 @@
+import logging
+
+logger = logging.getLogger()
+logger.setLevel("DEBUG")
+file_handler = logging.FileHandler("./log.txt", mode="a", encoding="utf-8")
+file_handler.setLevel("DEBUG")
+file_handler.setFormatter(
+    logging.Formatter(fmt="%(lineno)s---%(asctime)s---%(message)s")
+)
+logger.addHandler(file_handler)
+
 from copy import deepcopy
 
 import matplotlib.dates as mdates
@@ -36,7 +47,7 @@ def backtest_stats(account_value, value_col_name="account_value"):
         transactions=None,
         turnover_denom="AGB",
     )
-    print(perf_stats_all)
+    logging.info(perf_stats_all)
     return perf_stats_all
 
 

test.py

@@ -1,3 +1,14 @@
+import logging
+
+logger = logging.getLogger()
+logger.setLevel("DEBUG")
+file_handler = logging.FileHandler("./log.txt", mode="a", encoding="utf-8")
+file_handler.setLevel("DEBUG")
+file_handler.setFormatter(
+    logging.Formatter(fmt="%(lineno)s---%(asctime)s---%(message)s")
+)
+logger.addHandler(file_handler)
+
 from agents.elegantrl_models import DRLAgent as DRLAgent_erl
 from agents.rllib_models import DRLAgent as DRLAgent_rllib
 from agents.stablebaselines3_models import DRLAgent as DRLAgent_sb3
@@ -41,7 +52,7 @@ def test(
     # elegantrl needs state dim, action dim and net dim
     net_dimension = kwargs.get("net_dimension", 2**7)
     cwd = kwargs.get("cwd", "./" + str(model_name))
-    print("price_array: ", len(price_array))
+    logging.info("price_array: ", len(price_array))
 
     if drl_lib == "elegantrl":
         episode_total_assets = DRLAgent_erl.DRL_prediction(

trade.py

@@ -1,8 +1,7 @@
-from test import test
-
 from meta.env_stock_trading.env_stock_papertrading import (
     AlpacaPaperTrading,
 )
+from test import test
 
 
 def trade(

train.py

@@ -1,3 +1,14 @@
+import logging
+
+logger = logging.getLogger()
+logger.setLevel("DEBUG")
+file_handler = logging.FileHandler("./log.txt", mode="a", encoding="utf-8")
+file_handler.setLevel("DEBUG")
+file_handler.setFormatter(
+    logging.Formatter(fmt="%(lineno)s---%(asctime)s---%(message)s")
+)
+logger.addHandler(file_handler)
+
 from agents.elegantrl_models import DRLAgent as DRLAgent_erl
 from agents.rllib_models import DRLAgent as DRLAgent_rllib
 from agents.stablebaselines3_models import DRLAgent as DRLAgent_sb3
@@ -88,9 +99,9 @@ def train(
             tb_log_name=model_name,
             total_timesteps=total_timesteps,
         )
-        print("Training finished!")
+        logging.info("Training finished!")
         trained_model.save(cwd)
-        print("Trained model saved in " + str(cwd))
+        logging.info("Trained model saved in " + str(cwd))
 
     else:
         raise ValueError("DRL library input is NOT supported. Please check.")

tutorials/1-Introduction/China_A_share_market_tushare.py

@@ -7,6 +7,17 @@
 
 display.set_matplotlib_formats("svg")
 
+import logging
+
+logger = logging.getLogger()
+logger.setLevel("DEBUG")
+file_handler = logging.FileHandler("./log.txt", mode="a", encoding="utf-8")
+file_handler.setLevel("DEBUG")
+file_handler.setFormatter(
+    logging.Formatter(fmt="%(lineno)s---%(asctime)s---%(message)s")
+)
+logger.addHandler(file_handler)
+
 from meta import config
 from meta.data_processor import DataProcessor
 from main import check_and_make_directories
@@ -42,7 +53,7 @@
 
 pd.options.display.max_columns = None
 
-print("ALL Modules have been imported!")
+logging.info("ALL Modules have been imported!")
 
 
 # %% md
@@ -125,32 +136,32 @@
 # add_technical_indicator
 p.add_technical_indicator(config.INDICATORS)
 p.clean_data()
-print(f"p.dataframe: {p.dataframe}")
+logging.info(f"p.dataframe: {p.dataframe}")
 
 # %% md
 
 ### Split traning dataset
 
 train = p.data_split(p.dataframe, TRAIN_START_DATE, TRAIN_END_DATE)
-print(f"len(train.tic.unique()): {len(train.tic.unique())}")
+logging.info(f"len(train.tic.unique()): {len(train.tic.unique())}")
 
 # %%
 
-print(f"train.tic.unique(): {train.tic.unique()}")
+logging.info(f"train.tic.unique(): {train.tic.unique()}")
 
 # %%
 
-print(f"train.head(): {train.head()}")
+logging.info(f"train.head(): {train.head()}")
 
 # %%
 
-print(f"train.shape: {train.shape}")
+logging.info(f"train.shape: {train.shape}")
 
 # %%
 
 stock_dimension = len(train.tic.unique())
 state_space = stock_dimension * (len(config.INDICATORS) + 2) + 1
-print(f"Stock Dimension: {stock_dimension}, State Space: {state_space}")
+logging.info(f"Stock Dimension: {stock_dimension}, State Space: {state_space}")
 
 # %% md
 
@@ -182,7 +193,7 @@
 # %%
 
 env_train, _ = e_train_gym.get_sb_env()
-print(f"print(type(env_train)): {print(type(env_train))}")
+logging.info(f"print(type(env_train)): {logging.info(type(env_train))}")
 
 # %%
 
@@ -251,7 +262,7 @@
 # %%
 
 df_actions.to_csv("action.csv", index=False)
-print(f"df_actions: {df_actions}")
+logging.info(f"df_actions: {df_actions}")
 
 # %% md
 
@@ -298,8 +309,8 @@
     transactions=None,
     turnover_denom="AGB",
 )
-print("==============DRL Strategy Stats===========")
-print(f"perf_stats_all: {perf_stats_all}")
+logging.info("==============DRL Strategy Stats===========")
+logging.info(f"perf_stats_all: {perf_stats_all}")
 
 # %%
 
@@ -317,6 +328,6 @@
     transactions=None,
     turnover_denom="AGB",
 )
-print("==============Baseline Strategy Stats===========")
+logging.info("==============Baseline Strategy Stats===========")
 
-print(f"perf_stats_all: {perf_stats_all}")
+logging.info(f"perf_stats_all: {perf_stats_all}")

tutorials/2-Advance/Crypto_Feature_Importance.py

@@ -2,6 +2,17 @@
 
 warnings.filterwarnings("ignore")
 
+import logging
+
+logger = logging.getLogger()
+logger.setLevel("DEBUG")
+file_handler = logging.FileHandler("./log.txt", mode="a", encoding="utf-8")
+file_handler.setLevel("DEBUG")
+file_handler.setFormatter(
+    logging.Formatter(fmt="%(lineno)s---%(asctime)s---%(message)s")
+)
+logger.addHandler(file_handler)
+
 import pandas as pd
 import numpy as np
 from IPython import display
@@ -56,7 +67,7 @@
 
 pd.options.display.max_columns = None
 
-print("ALL Modules have been imported!")
+logging.info("ALL Modules have been imported!")
 
 # %% md
 
@@ -123,7 +134,7 @@
 p.download_data(ticker_list=ticker_list)
 p.clean_data()
 df = p.dataframe
-print(f"p.dataframe: {p.dataframe}")
+logging.info(f"p.dataframe: {p.dataframe}")
 
 
 def add_technical_indicator(df, tech_indicator_list):
@@ -146,12 +157,12 @@ def add_technical_indicator(df, tech_indicator_list):
 
 
 processed_df = add_technical_indicator(df, technical_indicator_list)
-print(f"processed_df: {processed_df.head()}")
+logging.info(f"processed_df: {processed_df.head()}")
 
 # Drop unecessary columns and make time as index
 processed_df.index = pd.to_datetime(processed_df.time)
 processed_df.drop("time", inplace=True, axis=1)
-print(processed_df.tail(20))
+logging.info(processed_df.tail(20))
 
 
 # IMPORTANT: Make sure that pd.Timedelta() is according to the time_interval to get the volatility for that time interval
@@ -174,7 +185,7 @@ def get_vol(prices, span=100):
 
 
 data_ohlcv = processed_df.assign(volatility=get_vol(processed_df.close)).dropna()
-print("data_ohlcv: ", data_ohlcv.head())
+logging.info("data_ohlcv: ", data_ohlcv.head())
 
 ##Adding Path Dependency: Triple-Barrier Method
 ###The labeling schema is defined as follows:
@@ -247,7 +258,7 @@ def get_barriers(
 
 
 barriers = get_barriers()
-print("barriers: ", barriers.head())
+logging.info("barriers: ", barriers.head())
 
 ####Function to get label for the dataset (0, 1, 2)
 # 0: hit the stoploss
@@ -306,7 +317,7 @@ def get_labels():
 
 
 get_labels()
-print("barriers after labeling: ", barriers.head())
+logging.info("barriers after labeling: ", barriers.head())
 # Merge the barriers with the main dataset and drop the last t_final + 1 barriers (as they are too close to the end)
 
 data_ohlcv = data_ohlcv.merge(
@@ -315,9 +326,9 @@ def get_labels():
     right_on="time",
 )
 data_ohlcv.drop(data_ohlcv.tail(t_final + 1).index, inplace=True)
-print("data_ohlcv after labeling: ", data_ohlcv.head())
+logging.info("data_ohlcv after labeling: ", data_ohlcv.head())
 # Count barrier hits ( 0 = stoploss, 1 = timeout, 2 = profit take)
-print(pd.Series(data_ohlcv["label_barrier"]).value_counts())
+logging.info(pd.Series(data_ohlcv["label_barrier"]).value_counts())
 
 ###Copying the Neural Network function from AI4Finance's ActorPPO agent
 # https://github.com/AI4Finance-Foundation/ElegantRL/blob/master/elegantrl/agents/net.py
@@ -365,7 +376,7 @@ def forward(self, x):
 
 
 model_NN1 = Net()
-print(model_NN1)
+logging.info(model_NN1)
 
 
 class ClassifierDataset(Dataset):
@@ -435,7 +446,7 @@ def __len__(self):
 test_loader = DataLoader(dataset=test_dataset, batch_size=1)
 # Check GPU
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
-print(device)
+logging.info(device)
 
 # Set optimizer
 criterion = nn.CrossEntropyLoss()
@@ -457,7 +468,7 @@ def train(fold, model, device, trainloader, optimizer, epoch):
         optimizer.step()
 
         if batch_idx % 100 == 0:
-            print(
+            logging.info(
                 "Train Fold/Epoch: {}/{} [{}/{} ({:.0f}%)]\ttrain_loss: {:.6f}".format(
                     fold,
                     epoch,
@@ -500,7 +511,7 @@ def test(fold, model, device, test_loader, correct_train, train_loss):
 
     # Print train accuracy for epoch
     # TODO: still a bug in summed up batch train loss
-    print(
+    logging.info(
         "\nTrain set for fold {}: Average train_loss: {:.4f}, Accuracy: {}/{} ({:.5f}%)".format(
             fold,
             train_loss,
@@ -511,7 +522,7 @@ def test(fold, model, device, test_loader, correct_train, train_loss):
     )
 
     # Print test result for epoch
-    print(
+    logging.info(
         "Test set for fold {}:  Average test_loss:  {:.4f}, Accuracy: {}/{} ({:.5f}%)\n".format(
             fold,
             test_loss,
@@ -559,11 +570,13 @@ def test(fold, model, device, test_loader, correct_train, train_loss):
     y_pred_nn1 = y_pred_nn1.cpu().detach().numpy()
 
 # print predction values
-print("labels in prediction:", np.unique(y_pred_nn1), "\n")
+logging.info("labels in prediction:", np.unique(y_pred_nn1), "\n")
 
 # print report
 label_names = ["long", "no bet", "short"]
-print(classification_report(y_test.astype(int), y_pred_nn1, target_names=label_names))
+logging.info(
+    classification_report(y_test.astype(int), y_pred_nn1, target_names=label_names)
+)
 
 
 def perturbation_rank(model, x, y, names):
@@ -608,4 +621,4 @@ def perturbation_rank(model, x, y, names):
     names,
 )
 
-print(display(rank))
+logging.info(display(rank))