upstage-llama-30b-instruct-2048-GPTQ / quantize_config.json
TheBloke's picture
Update for Transformers GPTQ support
71b5549
raw
history blame
186 Bytes
{
"bits": 4,
"group_size": -1,
"damp_percent": 0.01,
"desc_act": true,
"sym": true,
"true_sequential": true,
"model_name_or_path": null,
"model_file_base_name": "model"
}