try reload all

config/ppo_transfer/3DBall.yaml

       predict_return: true
       use_bisim: false
       separate_value_train: true
-      separate_value_net: true
       in_batch_alter: true
     network_settings:
       normalize: true

config/ppo_transfer/3DBallHard.yaml

       predict_return: true
       use_bisim: false
       separate_value_train: true
-      separate_value_net: true
       in_batch_alter: true
     network_settings:
       normalize: true

config/ppo_transfer/3DBallHardTransfer.yaml

       predict_return: true
       use_bisim: false
       separate_value_train: true
-      separate_value_net: true
-      load_policy: false
+      load_policy: true
-      load_value: false
+      load_value: true
       transfer_path: "results/3d-tr-s1/3DBall"
       use_transfer: true
       in_batch_alter: true

ml-agents/mlagents/trainers/policy/transfer_policy.py

         :param encoded_state: Tensor corresponding to encoded current state.
         :param encoded_next_state: Tensor corresponding to encoded next state.
         """
+        if not self.transfer:
+            encoded_state = tf.stop_gradient(encoded_state)
+
-        #if not self.transfer:
-        #    hidden = tf.stop_gradient(hidden)
-
+        
         for i in range(forward_layers):
             hidden = tf.layers.dense(
                 hidden,
         forward_layers: int,
         separate_train: bool = False,
     ):
+        if not self.transfer:
+            encoded_state = tf.stop_gradient(encoded_state)
-
         #if not self.transfer:
         #    hidden = tf.stop_gradient(hidden)