Sheep training flock of 10 fix?

2026-04-24 17:31:11 +01:00
parent c2da9c10e4
commit 4d7f365358
3 changed files with 4 additions and 18 deletions
@@ -104,7 +104,7 @@ def train_stage(n_sheep, steps, n_envs=4, prev_model=None, prev_vecnorm=None):
        model = PPO(
            "MlpPolicy", vn,
            learning_rate=3e-4, n_steps=2048, batch_size=256, n_epochs=10,
-            gamma=0.995, gae_lambda=0.95, clip_range=0.2, ent_coef=0.005,
+            gamma=0.995, gae_lambda=0.95, clip_range=0.2, ent_coef=0.02,
            vf_coef=0.5, max_grad_norm=0.5,
            policy_kwargs=dict(net_arch=[256, 256]),
            verbose=1,