xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-25 15:11:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc][DP] Guard mxfp4 implementation selection (#27484)

Browse Source 
Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
This commit is contained in:
Varun Sundar Rabindranath 2025-10-24 19:29:24 -04:00 committed by GitHub
parent 52efc34ebf
commit 269c4db0a4
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/model_executor/layers/quantization/mxfp4.py
View File

@ -794,7 +794,8 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
) )
else: else:
raise NotImplementedError( raise NotImplementedError(
"Incompatible Mxfp4 backend for EP batched experts format" f"Incompatible Mxfp4 backend ({self.mxfp4_backend}) for "
"EP batched experts format"
) )
else: else:
assert self.moe_quant_config is not None assert self.moe_quant_config is not None
@ -813,8 +814,12 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
return TrtLlmGenExperts(self.moe, self.moe_quant_config, **kwargs) return TrtLlmGenExperts(self.moe, self.moe_quant_config, **kwargs)
elif self.mxfp4_backend == Mxfp4Backend.MARLIN: elif self.mxfp4_backend == Mxfp4Backend.MARLIN:
return MarlinExperts(self.moe_quant_config) return MarlinExperts(self.moe_quant_config)
else: elif self.mxfp4_backend == Mxfp4Backend.TRITON:
return OAITritonExperts(self.moe_quant_config) return OAITritonExperts(self.moe_quant_config)
else:
raise NotImplementedError(
f"Incompatible Mxfp4 backend ({self.mxfp4_backend}) for EP"
)
def _route_and_experts( def _route_and_experts(
self, self,