{
    "dataset_dir": "/root/persona/datasets/finance-instruct-iter2/sft_train",
    "val_data_dir": "/root/persona/datasets/finance-instruct-iter2/sft_val",
    "training_data_size": 57337,
    "val_data_size": 6371,
    "lora_r": 64,
    "lora_alpha": 64,
    "target_modules": "{'up_proj', 'k_proj', 'down_proj', 'gate_proj', 'q_proj', 'o_proj', 'v_proj'}",
    "bias": "none",
    "task_type": "CAUSAL_LM",
    "output_dir": "/root/persona/training_results/finance-instruct-iter2/qwen3-4b-finance_instruct_iter2-20260313_231627",
    "num_train_epochs": 3,
    "per_device_train_batch_size": 2,
    "learning_rate": 1e-06,
    "lr_scheduler_type": "cosine",
    "warmup_ratio": 0.03,
    "bf16": true,
    "gradient_checkpointing": true,
    "gradient_checkpointing_kwargs": {
        "use_reentrant": false
    },
    "max_length": 4096,
    "logging_steps": 10,
    "save_steps": 200,
    "save_total_limit": 3,
    "optim": "adamw_torch",
    "report_to": "none",
    "run_name": null,
    "seed": 42,
    "ddp_find_unused_parameters": false,
    "dataloader_num_workers": 4,
    "model_init_kwargs": {
        "attn_implementation": "sdpa",
        "torch_dtype": "torch.bfloat16",
        "trust_remote_code": true
    },
    "eval_strategy": "steps",
    "eval_steps": 200
}