ppo_checkpoints/train_config.json

{
  "episode_dir": "./recorded_episodes",
  "output_dir": "./ppo_checkpoints_v3",
  "device": "cuda:0",
  "total_steps": 3000000,
  "n_rollout": 2048,
  "n_epochs": 2,
  "batch_size": 256,
  "lr": 5e-05,
  "gamma": 0.99,
  "gae_lambda": 0.95,
  "clip_eps": 0.2,
  "value_coef": 0.25,
  "entropy_coef": 0.05,
  "max_grad_norm": 0.5,
  "save_every": 50,
  "resume_from": "./ppo_checkpoints_v2/policy_update_00850.pt",
  "seed": 42
}