larger feature size

config/sac_transfer/Reacher.yaml

       reward_signal_steps_per_update: 20.0
       encoder_layers: 1
       policy_layers: 2
-      forward_layers: 2
+      forward_layers: 0
-      feature_size: 64
-      action_feature_size: 16
+      feature_size: 128
+      action_feature_size: 32
       separate_policy_train: true
       separate_policy_net: true
       separate_model_train: true

config/sac_transfer/ReacherTransfer.yaml

       reward_signal_steps_per_update: 20.0
       encoder_layers: 1
       policy_layers: 2
-      forward_layers: 2
+      forward_layers: 0
-      feature_size: 64
-      action_feature_size: 16
+      feature_size: 128
+      action_feature_size: 32
       separate_policy_train: true
       separate_policy_net: true
       reuse_encoder: false

ml-agents/mlagents/trainers/sac_transfer/optimizer.py

                 )
                 self.model_learning_rate = ModelUtils.create_schedule(
                     hyperparameters.model_schedule,
-                    lr,
+                    lr * 10,
                     self.policy.global_step,
                     int(max_step),
                     min_value=1e-10,