glm-4-9b-chat-int4-sym-inc / quantization_config.json
sys-lpot-val
add files
e1f1fbe
raw
history blame
575 Bytes
{
"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": false,
"quant_block_list": null,
"enable_norm_bias_tuning": false,
"dataset": "NeelNanda/pile-10k",
"autoround_version": "0.4.0.dev",
"quant_method": "intel/auto-round",
"backend": "auto_round:gptq:exllamav2"
}