use amp strategy

rl_game/get_up/config/ppo_cfg.yaml

@@ -17,7 +17,7 @@ params:
           name: default
         sigma_init:
           name: const_initializer
-          val: 0.80
+          val: 0.42
         fixed_sigma: False
     mlp:
       units: [512, 256, 128]
@@ -41,7 +41,7 @@ params:
     normalize_advantage: True
     gamma: 0.99
     tau: 0.95
-    learning_rate: 3e-4
+    learning_rate: 1e-4
     lr_schedule: adaptive
     kl_threshold: 0.013
     score_to_win: 20000
@@ -49,7 +49,7 @@ params:
     save_best_after: 50
     save_frequency: 100
     grad_norm: 0.8
-    entropy_coef: 0.00008
+    entropy_coef: 0.00011
     truncate_grads: True
     bounds_loss_coef: 0.01
     e_clip: 0.2