change parameter

rl_game/get_up/config/ppo_cfg.yaml

@@ -17,8 +17,8 @@ params:
           name: default
         sigma_init:
           name: const_initializer
-          val: 0.8
-        fixed_sigma: True
+          val: 0.7
+        fixed_sigma: False
     mlp:
       units: [512, 256, 128]
       activation: relu
@@ -41,9 +41,9 @@ params:
     normalize_advantage: True
     gamma: 0.99
     tau: 0.95
-    learning_rate: 5e-4
+    learning_rate: 3e-4
     lr_schedule: adaptive
-    kl_threshold: 0.013
+    kl_threshold: 0.008
     score_to_win: 20000
     max_epochs: 500000
     save_best_after: 50
@@ -54,7 +54,7 @@ params:
     bounds_loss_coef: 0.001
     e_clip: 0.2
     horizon_length: 64
-    minibatch_size: 8192
-    mini_epochs: 4
-    critic_coef: 2
+    minibatch_size: 4096
+    mini_epochs: 5
+    critic_coef: 1
     clip_value: True