Qwen2.5-32B-Instruct-GPTQ-Marlin / quantize_config.json
GSJL's picture
Upload of AutoGPTQ quantized model
d92b4e8 verified
raw
history blame
281 Bytes
{
"bits": 4,
"group_size": 128,
"damp_percent": 0.1,
"desc_act": false,
"static_groups": false,
"sym": true,
"true_sequential": true,
"model_name_or_path": "../tmp-gptq",
"model_file_base_name": "model",
"quant_method": "gptq",
"checkpoint_format": "marlin"
}