xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-28 04:39:09 +08:00
Code Issues Packages Projects Releases Wiki Activity

sync with #30957

Browse Source 
Signed-off-by: Xingyu Liu <charlotteliu12x@gmail.com>
This commit is contained in:
Xingyu Liu 2025-12-22 22:23:35 -08:00
parent f7344c971c
commit 78d47494df
1 changed files with 12 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
vllm/transformers_utils/model_arch_config_convertor.py
View File

@ -150,12 +150,18 @@ class ModelArchConfigConvertorBase:
producer_name = quant_cfg.get("producer", {}).get("name")
if producer_name == "modelopt":
quant_algo = quant_cfg.get("quantization", {}).get("quant_algo")
if quant_algo == "FP8":
quant_cfg["quant_method"] = "modelopt"
elif quant_algo == "NVFP4":
quant_cfg["quant_method"] = "modelopt_fp4"
elif quant_algo is not None:
raise ValueError(f"Unknown ModelOpt quant algo: {quant_algo}")
if quant_algo is not None:
quant_algo_upper = str(quant_algo).upper()
if quant_algo_upper in {
"FP8",
"FP8_PER_CHANNEL_PER_TOKEN",
"FP8_PB_WO",
}:
quant_cfg["quant_method"] = "modelopt"
elif quant_algo_upper == "NVFP4":
quant_cfg["quant_method"] = "modelopt_fp4"
else:
raise ValueError(f"Unknown ModelOpt quant algo: {quant_algo}")
if quant_cfg is not None:
# Use the community standard 'quant_method'