Amend for standing

rl_game/get_up/config/ppo_cfg.yaml

@@ -17,7 +17,7 @@ params:
           name: default
         sigma_init:
           name: const_initializer
-          val: 0
+          val: 0.8
         fixed_sigma: True
     mlp:
       units: [512, 256, 128]
@@ -39,7 +39,7 @@ params:
     reward_shaper:
       scale_value: 1.0
     normalize_advantage: True
-    gamma: 0.96
+    gamma: 0.99
     tau: 0.95
     learning_rate: 5e-4
     lr_schedule: adaptive
@@ -49,12 +49,12 @@ params:
     save_best_after: 50
     save_frequency: 100
     grad_norm: 1.0
-    entropy_coef: 0.02
+    entropy_coef: 0.05
     truncate_grads: True
     bounds_loss_coef: 0.001
     e_clip: 0.2
-    horizon_length: 32
-    minibatch_size: 16384
+    horizon_length: 64
+    minibatch_size: 8192
     mini_epochs: 4
     critic_coef: 2
     clip_value: True