dpo_argument.json

{
    "model_name_or_path": "Qwen/Qwen2-7B",
    "train_dataset_path": "./data/train.jsonl",
    "dev_dataset_path": "./data/dev.jsonl",
    "output_dir": "./checkpoints/dpo_ckpts",
    "per_device_train_batch_size": 1,
    "gradient_accumulation_steps": 8,
    "per_device_eval_batch_size": 1,
    "num_train_epochs": 1,
    "max_steps": 100,
    "learning_rate": 1e-06,
    "warmup_steps": 10,
    "logging_steps": 1,
    "evaluation_strategy": "steps",
    "save_strategy": "steps",
    "eval_steps": 100,
    "save_steps": 500,
    "max_seq_len": 4096,
    "max_prompt_len": 2048,
    "bf16": true,
    "fp16_opt_level": "O2",
    "do_train": true,
    "do_eval": true,
    "disable_tqdm": true,
    "load_best_model_at_end": true,
    "tensor_parallel_degree": 4,
    "sharding": "stage1",
    "use_flash_attention": true,
    "flash_mask":false,
    "recompute": false,
    "recompute_granularity": "full",
    "benchmark": false,
    "unified_checkpoint": true,
    "autotuner_benchmark":false,
    "beta": 0.1,
    "loss_type": "sigmoid",
    "label_smoothing": 0.0
  }