32 lines
719 B
JSON
32 lines
719 B
JSON
|
{
|
||
|
|
"bits": 4,
|
||
|
|
"group_size": 128,
|
||
|
|
"sym": true,
|
||
|
|
"data_type": "int",
|
||
|
|
"enable_quanted_input": true,
|
||
|
|
"enable_minmax_tuning": true,
|
||
|
|
"seqlen": 2048,
|
||
|
|
"batch_size": 8,
|
||
|
|
"scale_dtype": "torch.float16",
|
||
|
|
"lr": 0.001,
|
||
|
|
"minmax_lr": 0.001,
|
||
|
|
"gradient_accumulate_steps": 1,
|
||
|
|
"iters": 1000,
|
||
|
|
"amp": true,
|
||
|
|
"nsamples": 512,
|
||
|
|
"low_gpu_mem_usage": true,
|
||
|
|
"to_quant_block_names": null,
|
||
|
|
"enable_norm_bias_tuning": false,
|
||
|
|
"act_bits": 16,
|
||
|
|
"act_group_size": 128,
|
||
|
|
"act_sym": true,
|
||
|
|
"act_dynamic": true,
|
||
|
|
"act_data_type": "int",
|
||
|
|
"super_bits": null,
|
||
|
|
"super_group_size": null,
|
||
|
|
"autoround_version": "0.5.1",
|
||
|
|
"quant_method": "gptq",
|
||
|
|
"desc_act": false,
|
||
|
|
"true_sequential": false,
|
||
|
|
"damp_percent": 0.01
|
||
|
|
}
|