fix mass 3dball

config/sac_transfer/3DBallMassTransferCloud.yaml

     hyperparameters:
       learning_rate: 0.0003
       learning_rate_schedule: linear
-      batch_size: 256
+      model_schedule: linear
+      batch_size: 64
-      init_entcoef: 1.0
+      init_entcoef: 0.5
       reward_signal_steps_per_update: 10.0
       encoder_layers: 2
       policy_layers: 0
       action_feature_size: 32
       # separate_value_net: true
       separate_policy_train: true
-      separate_model_train: true
+      separate_model_train: true
       reuse_encoder: true
       in_epoch_alter: false
       in_batch_alter: true
     time_horizon: 1000
     summary_freq: 12000
     threaded: true
-    
-# parameter_randomization:
-#   mass:
-#     sampler_type: uniform
-#     sampler_parameters:
-#       min_value: 2.0
-#       max_value: 2.0

config/sac_transfer/3DBallMassTransferCloud1.yaml

     hyperparameters:
       learning_rate: 0.0003
       learning_rate_schedule: linear
-      batch_size: 256
+      model_schedule: linear
+      batch_size: 64
-      init_entcoef: 1.0
+      init_entcoef: 0.5
       reward_signal_steps_per_update: 10.0
       encoder_layers: 2
       policy_layers: 0
       action_feature_size: 32
       # separate_value_net: true
       separate_policy_train: true
-      separate_model_train: true
+      separate_model_train: true
       reuse_encoder: true
       in_epoch_alter: false
       in_batch_alter: true
     time_horizon: 1000
     summary_freq: 12000
     threaded: true
-    
-# parameter_randomization:
-#   mass:
-#     sampler_type: uniform
-#     sampler_parameters:
-#       min_value: 2.0
-#       max_value: 2.0

config/sac_transfer/3DBallMassTransferCloud2.yaml

     hyperparameters:
       learning_rate: 0.0003
       learning_rate_schedule: linear
-      batch_size: 256
+      model_schedule: linear
+      batch_size: 64
-      init_entcoef: 1.0
+      init_entcoef: 0.5
       reward_signal_steps_per_update: 10.0
       encoder_layers: 2
       policy_layers: 0
       action_feature_size: 32
       # separate_value_net: true
       separate_policy_train: true
-      separate_model_train: true
+      separate_model_train: true
       reuse_encoder: true
       in_epoch_alter: false
       in_batch_alter: true
     time_horizon: 1000
     summary_freq: 12000
     threaded: true
-    
-# parameter_randomization:
-#   mass:
-#     sampler_type: uniform
-#     sampler_parameters:
-#       min_value: 2.0
-#       max_value: 2.0

config/sac_transfer/3DBallMassTransferCloud3.yaml

     hyperparameters:
       learning_rate: 0.0003
       learning_rate_schedule: linear
-      batch_size: 256
+      model_schedule: linear
+      batch_size: 64
-      init_entcoef: 1.0
+      init_entcoef: 0.5
       reward_signal_steps_per_update: 10.0
       encoder_layers: 2
       policy_layers: 0
       action_feature_size: 32
       # separate_value_net: true
       separate_policy_train: true
-      separate_model_train: true
+      separate_model_train: true
       reuse_encoder: true
       in_epoch_alter: false
       in_batch_alter: true
     time_horizon: 1000
     summary_freq: 12000
     threaded: true
-    
-# parameter_randomization:
-#   mass:
-#     sampler_type: uniform
-#     sampler_parameters:
-#       min_value: 2.0
-#       max_value: 2.0

config/sac_transfer/3DBallMassTransferCloud4.yaml

     hyperparameters:
       learning_rate: 0.0003
       learning_rate_schedule: linear
-      batch_size: 256
+      model_schedule: linear
+      batch_size: 64
-      init_entcoef: 1.0
+      init_entcoef: 0.5
       reward_signal_steps_per_update: 10.0
       encoder_layers: 2
       policy_layers: 0
       action_feature_size: 32
       # separate_value_net: true
       separate_policy_train: true
-      separate_model_train: true
+      separate_model_train: true
       reuse_encoder: true
       in_epoch_alter: false
       in_batch_alter: true
     time_horizon: 1000
     summary_freq: 12000
     threaded: true
-    
-# parameter_randomization:
-#   mass:
-#     sampler_type: uniform
-#     sampler_parameters:
-#       min_value: 2.0
-#       max_value: 2.0