configs

config/ppo_transfer/CrawlerStatic.yaml

       lambd: 0.95
       num_epoch: 3
       learning_rate_schedule: linear
-      model_schedule: linear
+      model_schedule: constant
       encoder_layers: 2
       policy_layers: 2
       forward_layers: 2
       in_epoch_alter: false
-      in_batch_alter: false
+      in_batch_alter: true
       use_op_buffer: false
       use_var_predict: true
       with_prior: false

config/ppo_transfer/TransferCrawlerStatic.yaml

       use_transfer: true
       load_model: true
       train_model: false
-      transfer_path: "results/csold/CrawlerStatic"
+      transfer_path: "results/csold-const/CrawlerStatic"
     network_settings:
       normalize: true
       hidden_units: 512

config/sac_transfer/3DBall.yaml

       policy_layers: 1
       forward_layers: 1
       value_layers: 2
-      feature_size: 16
+      feature_size: 32
       reuse_encoder: false
       in_epoch_alter: false
       in_batch_alter: true

config/sac_transfer/3DBallHard.yaml

       learning_rate: 0.0003
       learning_rate_schedule: linear
       batch_size: 256
-      buffer_size: 50000
+      buffer_size: 24000
       buffer_init_steps: 0
       tau: 0.005
       steps_per_update: 10.0
       policy_layers: 1
       forward_layers: 1
       value_layers: 2
-      feature_size: 16
+      feature_size: 32
       reuse_encoder: false
       in_epoch_alter: false
       in_batch_alter: true
         gamma: 0.99
         strength: 1.0
     keep_checkpoints: 5
-    max_steps: 500000
+    max_steps: 1000000
     time_horizon: 1000
     summary_freq: 12000
     threaded: true

config/sac_transfer/3DBallHardTransfer.yaml

       learning_rate: 0.0003
       learning_rate_schedule: linear
       batch_size: 256
-      buffer_size: 50000
+      buffer_size: 24000
       buffer_init_steps: 0
       tau: 0.005
       steps_per_update: 10.0
       policy_layers: 1
       forward_layers: 1
       value_layers: 2
-      feature_size: 16
+      feature_size: 32
       reuse_encoder: false
       in_epoch_alter: false
       in_batch_alter: true
       use_transfer: true
       load_model: true
       train_model: false
-      transfer_path: "results/sac_model_ball_sep_bisim/3DBall"
+      transfer_path: "results/sac_model_ball_sep_linear_f32/3DBall"
     network_settings:
       normalize: true
       hidden_units: 128
         gamma: 0.99
         strength: 1.0
     keep_checkpoints: 5
-    max_steps: 500000
+    max_steps: 1000000
     time_horizon: 1000
     summary_freq: 12000
     threaded: true