xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 07:15:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Log] Optimize Log for FP8MOE (#25709)

Browse Source 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye 2025-09-30 22:18:43 -04:00 committed by GitHub
parent a73f6491c8
commit 1244948885
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/model_executor/layers/quantization/fp8.py
View File

@ -467,7 +467,8 @@ class Fp8MoEMethod(FusedMoEMethodBase):
logger.info_once("DeepGemm disabled: FlashInfer MOE is"
" enabled.")
elif (is_deep_gemm_supported()):
logger.info_once("Using DeepGemm kernels for Fp8MoEMethod.")
logger.debug_once(
"DeepGemm kernels available for Fp8MoEMethod.")
self.allow_deep_gemm = True
else:
logger.warning_once(
@ -481,9 +482,8 @@ class Fp8MoEMethod(FusedMoEMethodBase):
elif (current_platform.is_cuda()
and current_platform.is_device_capability(100)
and not self.flashinfer_moe_backend):
logger.info_once(
"Using CutlassBlockScaledGroupedGemm kernels for Fp8 MOE "
"on SM100.")
logger.debug_once(
"CutlassBlockScaledGroupedGemm available for Fp8MoEMethod.")
self.allow_cutlass_block_scaled_grouped_gemm = True
def create_weights(self, layer: Module, num_experts: int, hidden_size: int,