Meta-Llama-3-8B-Instruct-GPTQ / quantize_config.json
OxxoCodes's picture
Push model and config
dda0b64 verified
{
"bits": 4,
"group_size": 128,
"damp_percent": 0.1,
"desc_act": false,
"static_groups": false,
"sym": true,
"true_sequential": true,
"model_name_or_path": "/scratch/nbrown9/llama-3-8b-4bit-gptq",
"model_file_base_name": "gptq_model-4bit-128g",
"is_marlin_format": false,
"quant_method": "gptq"
}