ready for cloud training

config/ppo_transfer/3DBall.yaml

     time_horizon: 1000
     summary_freq: 12000
     threaded: true
+    transfer: true

config/ppo_transfer/3DBallHard.yaml

       num_epoch: 3
       learning_rate_schedule: linear
       conv_thres: 1e-3
+      use_transfer: true
+      transfer_path: "results/single_ball/3DBall"
     network_settings:
       normalize: true
       hidden_units: 128

ml-agents/mlagents/trainers/learn.py

     if options.env_settings.seed == -1:
         run_seed = np.random.randint(0, 10000)
     run_training(run_seed, options)
-    os.system('mlagents-learn config/ppo/3DBallHard.yaml --run=test --env=envs/3dballhard --force')
+    if options.behaviors.transfer:
+        os.system('mlagents-learn config/ppo_transfer/3DBallHard.yaml --env=envs/3dballhard --num-envs=4 --force')
 
 
 def main():

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

 
             with tf.variable_scope("value"):
                 if policy.use_continuous_act:
-                    self._create_cc_critic(h_size, num_layers, vis_encode_type)
+                    self._create_cc_critic(h_size, hyperparameters.value_layers, vis_encode_type)
-                    self._create_dc_critic(h_size, num_layers, vis_encode_type)
+                    self._create_dc_critic(h_size, hyperparameters.value_layers, vis_encode_type)
             
             with tf.variable_scope("optimizer/"):
                 self.learning_rate = ModelUtils.create_schedule(

ml-agents/mlagents/trainers/settings.py

     # Network
     encoder_layers: int = 1
     policy_layers: int = 1
+    value_layers: int = 1
     forward_layers: int = 1
     inverse_layers: int = 1