| { | |
| "base_model": "/data/users/zhenyugao/sjd/last_ckpt/ckpt_model_single_2016_iter_0025000/hf_model", | |
| "tokenizer": "/data/users/zhenyugao/sjd/LLaMA-Factory/saves/dated_mining_2013_final_v4_financialv2", | |
| "data": "/data/users/zhenyugao/sjd/LLaMA-Factory/data/sft_mixed_pt10tok_2016_with_nyt2k.json", | |
| "loss": "model_native", | |
| "world_size": 1, | |
| "per_device_batch": 1, | |
| "gradient_accumulation": 32, | |
| "global_batch": 32, | |
| "learning_rate": 1e-05, | |
| "epochs": 3, | |
| "total_steps": 5319, | |
| "max_length": 2048, | |
| "seed": 42 | |
| } |