{
  "model_name_or_path": "Qwen/Qwen2.5-Math-7B",
  "dataset_name_or_path": "./data/gsm8k_distilled_en_sft",
  "output_dir": "./checkpoints/Qwen2.5-Math-7B/gsm8k_distilled/",
  "per_device_train_batch_size": 1,
  "gradient_accumulation_steps": 4,
  "per_device_eval_batch_size": 1,
  "eval_accumulation_steps": 16,
  "num_train_epochs": 3,
  "learning_rate": 3e-05,
  "warmup_steps": 30,
  "logging_steps": 1,
  "evaluation_strategy": "no",
  "save_strategy": "epoch",
  "src_length": 8192,
  "max_length": 16384,
  "bf16": true,
  "fp16_opt_level": "O2",
  "do_train": true,
  "do_eval": false,
  "eval_steps": 10,
  "disable_tqdm": true,
  "load_best_model_at_end": false,
  "eval_with_do_generation": false,
  "metric_for_best_model": "accuracy",
  "recompute": true,
  "save_total_limit": 1,
  "tensor_parallel_degree": 1,
  "pipeline_parallel_degree": 1,
  "sharding": "stage2",
  "zero_padding": true,
  "unified_checkpoint": true,
  "use_flash_attention": true,
  "flash_mask": true
}