training_config.json

{
  "mode": "codi_single",
  "model": "Qwen/Qwen3-1.7B",
  "rollouts": "data/rollouts.jsonl",
  "teacher_states": "data/teacher_states_v3.pt",
  "output_dir": "checkpoints/r11_single_perstep_g1",
  "epochs": 3,
  "batch_size": 2,
  "grad_accum": 8,
  "lr": 0.0002,
  "gamma": 1.0,
  "warmup_ratio": 0.03,
  "num_latent": 6,
  "max_prompt_len": 512,
  "max_answer_len": 128,
  "max_len": 1024,
  "answer_format": "full",
  "per_step_distill": true,
  "log_every": 10,
  "seed": 42,
  "lora_rank": 32,
  "lora_alpha": 16,
  "best_loss": 0.6426934776270491,
  "total_time_s": 1477.4382388591766
}