args.json

{
  "model_name": "Qwen/Qwen2-VL-7B-Instruct",
  "train_jsonl": "DPO/reasoning/Qwen2-VL-7B-Instruct_single__20250922_111950.json",
  "eval_jsonl": "",
  "output_dir": "./qwen2vl-dpo/20250922_191759",
  "epochs": 1,
  "batch_size": 1,
  "grad_accum_steps": 4,
  "lr": 1e-06,
  "weight_decay": 0.0,
  "warmup_ratio": 0.03,
  "beta": 0.1,
  "max_length": 1536,
  "max_grad_norm": 1.0,
  "logging_steps": 1,
  "save_steps": 0,
  "seed": 42,
  "fp16": false,
  "bf16": true,
  "clear_cache_every": 1
}