ppo conf

rl/train_ppo.yaml

 algo: PPO
 seed: 42
+torch_deterministic: true
+
 model:
   hidden_sizes: [64,64]
   use_lstm: False
   act_limit: 1.0
+
 params:
   learning_rate: 0.0003
-  gamma: 0.99
-  anneal_lr: True
-  clip_coef: 0.2
-  ent_coef: 0.2
-  vf_coef: 0.5
-  gae: True
+  anneal_lr: true
+  ent_coef: 0.0
+  gae: true
   gae_lambda: 0.95
-  norm_adv: True
+  gamma: 0.99
   max_grad_norm: 0.5
-  target_kl: 0.015 #0.015
-  clip_vloss: True
+  norm_adv: true
+  target_kl: null
+  vf_coef: 0.5
+  clip_coef: 0.2
+  clip_vloss: true
 
 env:
   path: ../simulations/rc_car_foyer_track
   num_envs: 4
   num_stacked_obs: 4
   num_features: 2
- 
+
 train:
   total_steps: 2000000
-  num_steps: 2048
+  num_steps: 512
   num_minibatches: 32
   update_epochs: 10
-  eval_every: 10
+  eval_every: 1
   num_eval_episodes: 5
   save_freq: 10