| { |
| "task": "objects", |
| "model_name_or_path": "/data1/wenbiao_zhao/models/Qwen3-8B", |
| "train_jsonl": "data/formal/v6/objects/train.jsonl", |
| "output_dir": "runs/formal/v6-8b/objects", |
| "eval_jsonl": "data/formal/v6/objects/dev.jsonl", |
| "model_revision": "b968826d9c46dd6066d109eabc6255188de91218", |
| "tokenizer_name_or_path": null, |
| "max_steps": 900, |
| "max_length": 2048, |
| "per_device_train_batch_size": 1, |
| "gradient_accumulation_steps": 4, |
| "learning_rate": 0.0001, |
| "lr_scheduler_type": "cosine", |
| "warmup_steps": 10, |
| "lora_r": 16, |
| "lora_alpha": 32, |
| "lora_dropout": 0.0, |
| "lora_target_modules": "all-linear", |
| "seed": 20260919, |
| "bf16": true, |
| "fp16": false, |
| "gradient_checkpointing": true, |
| "logging_steps": 10, |
| "save_steps": 150, |
| "save_total_limit": 3, |
| "packing": false, |
| "completion_only_loss": true, |
| "local_files_only": true, |
| "allow_download": false, |
| "max_samples": null, |
| "eval_fraction": 0.0, |
| "dev_exact_match_every": 150, |
| "dev_exact_match_samples": 24, |
| "dev_exact_match_max_new_tokens": 96, |
| "report_to": "none", |
| "dataset_num_proc": 1, |
| "extra": {} |
| } |
|
|
