minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[RL] Support FlashInfer TRT-LLM NVFP4 MoE in the RL weight checker (#31085)

Browse Source 
Signed-off-by: zhihaow6 <zhihaow6@illinois.edu>
This commit is contained in:
Zhihao Wang
2026-07-24 16:01:54 -07:00
committed by GitHub
parent 0a212c6119
commit 14d6e1d3b1
1 changed files with 7 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/utils/weight_checker_comparator.py
+7 -1
View File
@@ -161,7 +161,13 @@ def select_comparable_weight(quant_method) -> Optional[type]:
and not quant_method.use_mxfp8 and not quant_method.use_mxfp8
): ):
return Fp8BlockComparable return Fp8BlockComparable
if isinstance(quant_method, (ModelOptFp4LinearMethod, ModelOptNvFp4FusedMoEMethod)): if isinstance(quant_method, ModelOptNvFp4FusedMoEMethod):
if getattr(quant_method, "enable_flashinfer_trtllm_moe", False):
return None
raise NotImplementedError(
f"weight checker has no ComparableWeight for {type(quant_method).__name__}"
)
if isinstance(quant_method, ModelOptFp4LinearMethod):
raise NotImplementedError( raise NotImplementedError(
f"weight checker has no ComparableWeight for {type(quant_method).__name__}" f"weight checker has no ComparableWeight for {type(quant_method).__name__}"
) )