Add action scaler for training continuous action-space dataset

Author: takuseno

minerva/tasks.py

@@ -52,6 +52,10 @@ def train(algo_name,
         augmentation = DrQPipeline(augmentations, n_mean=n_mean)
         params['augmentation'] = augmentation
 
+    # add action scaler if continuous action-space
+    if not dataset.is_action_discrete():
+        params['action_scaler'] = 'min_max'
+
     # train
     algo = create_algo(algo_name, dataset.is_action_discrete(), **params)
     algo.fit(train_data,