量化

sglang v0.5.1 版本上quantation gptq 不再依赖 vllm 模块 https://github.com/sgl-project/sglang/pull/9191

import sys
from transformers import AutoConfig

model=sys.argv[1]
print(model)
config = AutoConfig.from_pretrained(model)

print(config)

config中可以看到：

"quantization_config": {
    "batch_size": 1,
    "bits": 4,
    "block_name_to_quantize": null,
    "cache_block_outputs": true,
    "damp_percent": 0.1,
    "dataset": null,
    "desc_act": false,
    "exllama_config": {
      "version": 1
    },
    "group_size": 128,
    "max_input_length": null,
    "model_seqlen": null,
    "module_name_preceding_first_block": null,
    "modules_in_block_to_quantize": null,
    "pad_token_id": null,
    "quant_method": "gptq",
    "sym": true,
    "tokenizer": null,
    "true_sequential": true,
    "use_cuda_fp16": false,
    "use_exllama": true
  },