feat(PPOTrainer): decentralized message passing babyyyy

2026-05-02 21:50:42 +02:00 · 2026-05-02 21:50:42 +02:00 · 6114e36e27
commit 6114e36e27
parent 9398592e34
6 changed files with 122 additions and 34 deletions
--- a/configs/ppo/chickendinnerwinner.yaml
+++ b/configs/ppo/chickendinnerwinner.yaml
@ -0,0 +1,16 @@
+anneal_lr: true
+clip_coef: 0.2
+clip_vloss: true
+ent_coef: 0.001
+gae_lambda: 0.95
+gamma: 0.99
+learning_rate: 0.0001
+max_grad_norm: 0.5
+norm_adv: true
+num_envs: 32
+num_minibatches: 32
+num_steps: 64
+target_kl: 0.02
+total_timesteps: 12288000
+update_epochs: 4
+vf_coef: 1.0