phi-2-int4-inc / quantize_config.json
wenhuach's picture
replace with sym quantization as there is a large accuracy drop due to kernel issue
f201b74
{
"bits": 4,
"group_size": 128,
"damp_percent": 0.01,
"desc_act": false,
"static_groups": false,
"sym": true,
"true_sequential": false,
"model_name_or_path": null,
"model_file_base_name": "model",
"is_marlin_format": false,
"quant_method": "intel/auto-round",
"autoround_version": "0.2.0.dev",
"iters": 1000,
"lr": 0.001,
"minmax_lr": 0.001,
"enable_minmax_tuning": true,
"enable_quanted_input": true,
"scale_dtype": "float16"
}