checkpoint-epoch-2/grpo_config.json

{
  "beta": 0.05,
  "gamma": 0.99,
  "clip_range": 0.2,
  "learning_rate": 5e-06,
  "batch_size": 1,
  "gradient_accumulation_steps": 4,
  "max_epochs": 2,
  "max_length": 256,
  "max_new_tokens": 100,
  "temperature": 0.8,
  "top_p": 0.9,
  "do_sample": true,
  "weight_decay": 0.01,
  "warmup_ratio": 0.1,
  "max_grad_norm": 1.0,
  "force_gpu": true,
  "mixed_precision": false,
  "gradient_checkpointing": true,
  "max_memory_per_gpu": "14GB",
  "logging_steps": 5,
  "eval_steps": 50,
  "save_steps": 50,
  "save_total_limit": 3,
  "reward_method": "adaptive",
  "baseline_type": "batch_mean",
  "ema_alpha": 0.1,
  "kl_target": 0.01,
  "kl_horizon": 10000,
  "adaptive_kl": true
}