{
  "sft_checkpoint": "./sft_checkpoint",
  "rollouts_per_difficulty": 16,
  "difficulty_mix": [
    1,
    1,
    2,
    2,
    3
  ],
  "num_train_epochs": 10.0,
  "per_device_batch": 12,
  "grad_accum": 4,
  "num_generations": 8,
  "lr": 1e-06,
  "beta": 0.05,
  "gamma": 0.98,
  "seed": 42,
  "n_unique_prompts": 247,
  "n_state_snapshots": 612,
  "use_vllm": false
}