| { |
| "adapter_path": "adapters/qwen_1.5b_nogai", |
| "batch_size": 2, |
| "clear_cache_threshold": 0, |
| "config": null, |
| "data": "./nogai_data", |
| "fine_tune_type": "lora", |
| "grad_accumulation_steps": 1, |
| "grad_checkpoint": false, |
| "iters": 2500, |
| "learning_rate": 0.0002, |
| "lora_parameters": { |
| "rank": 8, |
| "dropout": 0.0, |
| "scale": 20.0 |
| }, |
| "lr_schedule": null, |
| "mask_prompt": false, |
| "max_seq_length": 512, |
| "model": "./local_qwen_1.5B", |
| "num_layers": 16, |
| "optimizer": "adam", |
| "optimizer_config": { |
| "adam": {}, |
| "adamw": {}, |
| "muon": {}, |
| "sgd": {}, |
| "adafactor": {} |
| }, |
| "project_name": null, |
| "report_to": null, |
| "resume_adapter_file": null, |
| "save_every": 500, |
| "seed": 0, |
| "steps_per_eval": 200, |
| "steps_per_report": 10, |
| "test": false, |
| "test_batches": 500, |
| "train": true, |
| "val_batches": 25 |
| } |
