| { | |
| "model_type": "causal_lm", | |
| "base_model": "meta-llama/Llama-3.3-70B-Instruct", | |
| "quantization": { | |
| "load_in_8bit": true, | |
| "quantization_type": "standard_8bit", | |
| "compute_dtype": "bfloat16" | |
| }, | |
| "fine_tuning": { | |
| "method": "PEFT", | |
| "technique": "LoRA", | |
| "lora_rank": 16, | |
| "lora_alpha": 16, | |
| "target_modules": [ | |
| "qproj", | |
| "kproj", | |
| "vproj", | |
| "oproj", | |
| "gateproj", | |
| "upproj", | |
| "downproj" | |
| ], | |
| "dropout": 0.05 | |
| }, | |
| "device_map": "auto", | |
| "trust_remote_code": true, | |
| "tokenizer_options": { | |
| "use_fast": true | |
| } | |
| } | |