Merge pull request #1 from TunnRL/code_review

making code platform independent + fixing environment problem

Author: geograz

README.md

@@ -14,7 +14,19 @@ DOI: XXXXXXXXXXXXXXXXXXXXXXXX
 
 ## Requirements and folder structure
 
-Use the `requirements.txt` file to download the required packages to run the code. We recommend using a package management system like conda for this purpose.
+Use the `requirements.txt` or `environment.yaml` file to download the required packages to run the code. We recommend using a package management system like conda or pipenv for this purpose. Using conda:
+
+1. Create an environment called `rl_cutter` using `environment.yaml` with the help of `conda`. If you get pip errors, install pip libraries manually, e.g. `pip install pandas`
+
+   ```bash
+   conda env create --file environment.yaml
+   ```
+
+2. Activate the new environment with:
+
+   ```bash
+   conda activate rl_cutter
+   ```
 
 The code framework depends on a certain folder structure. The python files should be placed in the main directory. The set up should be done in the following way:
 ```
@@ -29,5 +41,5 @@ TunnRL_TBM_maintenance
 ```
 Either set up the folder structure manually or on Linux run:
 ```bash
-bash folder_structure.sh
+bash make_folder_structure.sh
 ```

environment.yaml

@@ -0,0 +1,16 @@
+name: rl_cutter
+channels:
+- conda-forge
+- defaults
+dependencies:
+- python=3.9.7
+- joblib=1.1.0
+- matplotlib=3.5.1
+- numpy
+- pandas
+- rich
+- pip
+- pip:
+  - optuna==2.10.0
+  - tensorforce
+  - keras==2.6.0

make_folder_structure.sh

@@ -0,0 +1 @@
+mkdir results checkpoints graphics

src/A_main.py

@@ -16,11 +16,14 @@
 import numpy as np
 import optuna
 import pandas as pd
+from pathlib import Path
 from tensorforce.agents import Agent
 from tensorforce.environments import Environment
+from rich.traceback import install
 
 from XX_maintenance_lib import maintenance, CustomEnv, plotter
 
+install() # for better error messages
 
 ###############################################################################
 # Constants and fixed variables
@@ -176,7 +179,7 @@ def objective(trial):
 
 if OPTIMIZATION is True:  # study
     try:  # evtl. load already existing study if one exists
-        study = joblib.load(fr'results\{STUDY}.pkl')
+        study = joblib.load(Path(f'results/{STUDY}.pkl'))
         print('prev. study loaded')
     except FileNotFoundError:  # or create a new study
         study = optuna.create_study(direction='maximize')
@@ -185,10 +188,10 @@ def objective(trial):
     # saved and can be checked every 2 trials
     for _ in range(200):  # save every second study
         study.optimize(objective, n_trials=2)
-        joblib.dump(study, fr"results\{STUDY}.pkl")
+        joblib.dump(study, Path(f"results/{STUDY}.pkl"))
 
         df = study.trials_dataframe()
-        df.to_csv(fr'results\{STUDY}.csv')
+        df.to_csv(Path(f'results/{STUDY}.csv'))
     # print results of study
     trial = study.best_trial
     print('\nhighest reward: {}'.format(trial.value))

src/B_optimization_analyzer.py

@@ -14,6 +14,7 @@
 import matplotlib.pyplot as plt
 import numpy as np
 import pandas as pd
+from pathlib import Path
 
 
 ###############################################################################
@@ -25,8 +26,8 @@
 # processing
 
 # load data from completed OPTUNA study
-STUDY = fr"results\{name}.pkl"  # name of the saved study file
-DF = fr'results\{name}.csv'  # name of the csv. where logs from the study are
+STUDY = Path(f"results/{name}.pkl")  # name of the saved study file
+DF = Path(f'results\{name}.csv')  # name of the csv. where logs from the study are
 # load data
 df_study = pd.read_csv(DF)
 study = joblib.load(STUDY)
@@ -61,7 +62,7 @@
 ax.set_xlabel('trial number')
 ax.set_ylabel('reward')
 plt.tight_layout()
-plt.savefig(fr'graphics\{name}_optimization_progress.svg')
+plt.savefig(Path(f'graphics/{name}_optimization_progress.svg'))
 plt.close()
 
 # scatterplot of indivdual hyperparameters vs. reward
@@ -91,7 +92,7 @@
         ax.set_xscale('log')
 
 plt.tight_layout()
-plt.savefig(fr'graphics\{name}_optimization_scatter.svg')
+plt.savefig(Path(f'graphics/{name}_optimization_scatter.svg'))
 plt.close()
 
 # plot of the progress of individual runs
@@ -114,5 +115,5 @@
 ax.set_xlabel('episodes')
 ax.set_ylabel('reward')
 plt.tight_layout()
-plt.savefig(fr'graphics\{name}_optimization_intermediates.svg')
+plt.savefig(Path(f'graphics/{name}_optimization_intermediates.svg'))
 plt.close()

src/XX_maintenance_lib.py

@@ -15,6 +15,7 @@
 import matplotlib.pyplot as plt
 import matplotlib.gridspec as gridspec
 import numpy as np
+from pathlib import Path
 from tensorforce.environments import Environment
 
 
@@ -105,8 +106,8 @@ def sample_ep_plot(self, states, actions, rewards, ep, savename):
         ax.grid(alpha=0.5)
 
         plt.tight_layout()
-        plt.savefig(fr'checkpoints\{savename}_sample.png', dpi=600)
-        plt.savefig(fr'checkpoints\{savename}_sample.svg')
+        plt.savefig(Path(f'checkpoints/{savename}_sample.png'), dpi=600)
+        plt.savefig(Path(f'checkpoints/{savename}_sample.svg'))
         plt.close()
 
     def trainingprogress_plot(self, df, summed_actions, name):
@@ -142,8 +143,8 @@ def trainingprogress_plot(self, df, summed_actions, name):
         ax3.set_xlabel('episodes')
 
         plt.tight_layout()
-        plt.savefig(fr'checkpoints\{name}_progress.png', dpi=600)
-        plt.savefig(fr'checkpoints\{name}_progress.svg')
+        plt.savefig(Path(f'checkpoints/{name}_progress.png'), dpi=600)
+        plt.savefig(Path(f'checkpoints/{name}_progress.svg'))
         plt.close()
 
 
@@ -285,7 +286,7 @@ def save(self, AGENT, train_start, ep, states, actions, rewards, df,
         self.trainingprogress_plot(df, summed_actions, name)
 
         agent.save(directory='checkpoints', filename=name, format='hdf5')
-        df.to_csv(fr'checkpoints\{name}.csv', index=False)
+        df.to_csv(Path(f'checkpoints/{name}.csv', index=False))
 
 
 if __name__ == "__main__":
@@ -314,7 +315,7 @@ def save(self, AGENT, train_start, ep, states, actions, rewards, df,
     ax.set_xlabel('cutters to change')
     ax.set_ylabel('total maintenance time [min]')
     plt.tight_layout()
-    plt.savefig(r'graphics\cutter_changing_function.svg')
+    plt.savefig(Path('graphics/cutter_changing_function.svg'))
 
     ##########################################################################
     # 3D plot of reward functions
@@ -366,4 +367,4 @@ def save(self, AGENT, train_start, ep, states, actions, rewards, df,
     ax.set_zlabel('reward')
     ax.set_zlim(top=1)
 
-    plt.savefig(r'graphics\reward_function.svg')
+    plt.savefig(Path(f'graphics/reward_function.svg'))