test-dpo-3 / adapter_config.json
raniero's picture
submission (DPO)
20bea6a verified
{
"base_model": "mistralai/Mistral-7B-Instruct-v0.2",
"method": "LORA",
"task": "DPO",
"r": 4,
"lora_alpha": 8,
"lora_dropout": 0.05,
"target_modules": [
"q_proj",
"v_proj"
]
}