multiple sizes configs

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

                     train_vars = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES)
                 elif self.train_type == "encoding":
                     train_vars = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, "encoding")
+                elif self.train_type == "policy":
+                    train_vars = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, "encoding")
+                    train_vars += tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, "policy")
+                    train_vars += tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, "value")
                 print("trainable", train_vars)
                 # train_vars = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, "encoding")
                 # train_vars += tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, "policy")

ml-agents/mlagents/trainers/settings.py

 
     separate_value_train: bool = False
     separate_policy_train: bool = False
+    seprate_value_net: bool = False
     use_var_encoder: bool = False
     use_var_predict: bool = False
     with_prior: bool = False

config/ppo_transfer/CrawlerStatic128.yaml

+behaviors:
+  CrawlerStatic:
+    trainer_type: ppo_transfer
+    hyperparameters:
+      batch_size: 2024
+      buffer_size: 20240
+      learning_rate: 0.0003
+      beta: 0.005
+      epsilon: 0.2
+      lambd: 0.95
+      num_epoch: 3
+      learning_rate_schedule: linear
+      encoder_layers: 2
+      policy_layers: 2
+      value_layers: 2
+      feature_size: 128
+      reuse_encoder: true
+      in_epoch_alter: true
+    network_settings:
+      normalize: true
+      hidden_units: 512
+      num_layers: 3
+      vis_encode_type: simple
+    reward_signals:
+      extrinsic:
+        gamma: 0.995
+        strength: 1.0
+    keep_checkpoints: 5
+    max_steps: 10000000
+    time_horizon: 1000
+    summary_freq: 30000
+    threaded: true

config/ppo_transfer/CrawlerStatic128_256.yaml

+behaviors:
+  CrawlerStatic:
+    trainer_type: ppo_transfer
+    hyperparameters:
+      batch_size: 2024
+      buffer_size: 20240
+      learning_rate: 0.0003
+      beta: 0.005
+      epsilon: 0.2
+      lambd: 0.95
+      num_epoch: 3
+      learning_rate_schedule: linear
+      encoder_layers: 2
+      policy_layers: 2
+      value_layers: 2
+      feature_size: 128
+      reuse_encoder: true
+      in_epoch_alter: true
+    network_settings:
+      normalize: true
+      hidden_units: 256
+      num_layers: 3
+      vis_encode_type: simple
+    reward_signals:
+      extrinsic:
+        gamma: 0.995
+        strength: 1.0
+    keep_checkpoints: 5
+    max_steps: 10000000
+    time_horizon: 1000
+    summary_freq: 30000
+    threaded: true

config/ppo_transfer/CrawlerStatic256.yaml

+behaviors:
+  CrawlerStatic:
+    trainer_type: ppo_transfer
+    hyperparameters:
+      batch_size: 2024
+      buffer_size: 20240
+      learning_rate: 0.0003
+      beta: 0.005
+      epsilon: 0.2
+      lambd: 0.95
+      num_epoch: 3
+      learning_rate_schedule: linear
+      encoder_layers: 2
+      policy_layers: 2
+      value_layers: 2
+      feature_size: 256
+      reuse_encoder: true
+      in_epoch_alter: true
+    network_settings:
+      normalize: true
+      hidden_units: 512
+      num_layers: 3
+      vis_encode_type: simple
+    reward_signals:
+      extrinsic:
+        gamma: 0.995
+        strength: 1.0
+    keep_checkpoints: 5
+    max_steps: 10000000
+    time_horizon: 1000
+    summary_freq: 30000
+    threaded: true

config/ppo_transfer/CrawlerStatic256ppo.yaml

+behaviors:
+  CrawlerStatic:
+    trainer_type: ppo
+    hyperparameters:
+      batch_size: 2024
+      buffer_size: 20240
+      learning_rate: 0.0003
+      beta: 0.005
+      epsilon: 0.2
+      lambd: 0.95
+      num_epoch: 3
+      learning_rate_schedule: linear
+    network_settings:
+      normalize: true
+      hidden_units: 256
+      num_layers: 3
+      vis_encode_type: simple
+    reward_signals:
+      extrinsic:
+        gamma: 0.995
+        strength: 1.0
+    keep_checkpoints: 5
+    max_steps: 10000000
+    time_horizon: 1000
+    summary_freq: 30000
+    threaded: true

config/ppo_transfer/CrawlerStatic32_256.yaml

+behaviors:
+  CrawlerStatic:
+    trainer_type: ppo_transfer
+    hyperparameters:
+      batch_size: 2024
+      buffer_size: 20240
+      learning_rate: 0.0003
+      beta: 0.005
+      epsilon: 0.2
+      lambd: 0.95
+      num_epoch: 3
+      learning_rate_schedule: linear
+      encoder_layers: 2
+      policy_layers: 2
+      value_layers: 2
+      feature_size: 32
+      reuse_encoder: true
+      in_epoch_alter: true
+    network_settings:
+      normalize: true
+      hidden_units: 256
+      num_layers: 3
+      vis_encode_type: simple
+    reward_signals:
+      extrinsic:
+        gamma: 0.995
+        strength: 1.0
+    keep_checkpoints: 5
+    max_steps: 10000000
+    time_horizon: 1000
+    summary_freq: 30000
+    threaded: true

config/ppo_transfer/CrawlerStatic64_256.yaml

+behaviors:
+  CrawlerStatic:
+    trainer_type: ppo_transfer
+    hyperparameters:
+      batch_size: 2024
+      buffer_size: 20240
+      learning_rate: 0.0003
+      beta: 0.005
+      epsilon: 0.2
+      lambd: 0.95
+      num_epoch: 3
+      learning_rate_schedule: linear
+      encoder_layers: 2
+      policy_layers: 2
+      value_layers: 2
+      feature_size: 64
+      reuse_encoder: true
+      in_epoch_alter: true
+    network_settings:
+      normalize: true
+      hidden_units: 256
+      num_layers: 3
+      vis_encode_type: simple
+    reward_signals:
+      extrinsic:
+        gamma: 0.995
+        strength: 1.0
+    keep_checkpoints: 5
+    max_steps: 10000000
+    time_horizon: 1000
+    summary_freq: 30000
+    threaded: true