Skip to content

GitLab

Unverified Commit 12449488 authored by Wentao Ye's avatar Wentao Ye Committed by GitHub
Browse files

[Log] Optimize Log for FP8MOE (#25709) 


Signed-off-by: default avataryewentao256 <zhyanwentao@126.com>
parent a73f6491
Hide whitespace changes
Inline Side-by-side
Showing with 4 additions and 4 deletions
vllm/model_executor/layers/quantization/fp8.py
View file @ 12449488
...@@ -467,7 +467,8 @@ class Fp8MoEMethod(FusedMoEMethodBase): ...@@ -467,7 +467,8 @@ class Fp8MoEMethod(FusedMoEMethodBase):
logger.info_once("DeepGemm disabled: FlashInfer MOE is" logger.info_once("DeepGemm disabled: FlashInfer MOE is"
" enabled.") " enabled.")
elif (is_deep_gemm_supported()): 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 self.allow_deep_gemm = True
else: else:
logger.warning_once( logger.warning_once(
...@@ -481,9 +482,8 @@ class Fp8MoEMethod(FusedMoEMethodBase): ...@@ -481,9 +482,8 @@ class Fp8MoEMethod(FusedMoEMethodBase):
elif (current_platform.is_cuda() elif (current_platform.is_cuda()
and current_platform.is_device_capability(100) and current_platform.is_device_capability(100)
and not self.flashinfer_moe_backend): and not self.flashinfer_moe_backend):
logger.info_once( logger.debug_once(
"Using CutlassBlockScaledGroupedGemm kernels for Fp8 MOE " "CutlassBlockScaledGroupedGemm available for Fp8MoEMethod.")
"on SM100.")
self.allow_cutlass_block_scaled_grouped_gemm = True self.allow_cutlass_block_scaled_grouped_gemm = True
def create_weights(self, layer: Module, num_experts: int, hidden_size: int, def create_weights(self, layer: Module, num_experts: int, hidden_size: int,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment