| { | |
| "max_seq_len": 4096, | |
| "epochs": 3, | |
| "warmup_ratio": 0.03, | |
| "lora_dropout": 0.05, | |
| "seed": 42, | |
| "model_id": "Qwen/Qwen2.5-Coder-7B-Instruct", | |
| "target_modules": [ | |
| "q_proj", | |
| "k_proj", | |
| "v_proj", | |
| "o_proj", | |
| "gate_proj", | |
| "up_proj", | |
| "down_proj" | |
| ], | |
| "lora_r": 32, | |
| "lora_alpha": 64, | |
| "learning_rate": 0.0001, | |
| "per_device_bs": 2, | |
| "grad_accum": 8 | |
| } |