youngkingdom/vllm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

precommit

Browse Source 
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
This commit is contained in:
Tyler Michael Smith
2025-10-29 20:26:12 +00:00
parent 1af476b0e9
commit 7557a67655
1 changed files with 11 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
vllm/model_executor/layers/fused_moe/batched_triton_or_deep_gemm_moe.py
View File

@ -45,14 +45,17 @@ class BatchedTritonOrDeepGemmExperts(mk.FusedMoEPermuteExpertsUnpermute):
# If deep gemm was requested but is not available (either due to
# unsupported configuration or missing dependencies), check if
# we should allow fallback to batched triton kernel
if deep_gemm_requested and self.batched_deep_gemm_experts is None:
if not envs.VLLM_ALLOW_BATCHED_TRITON_FALLBACK:
raise RuntimeError(
"DeepGemm was requested but is not available. "
"The batched triton kernel fallback is disabled by default. "
"Set VLLM_ALLOW_BATCHED_TRITON_FALLBACK=1 to enable the fallback "
"for debugging purposes."
)
if (
deep_gemm_requested
and self.batched_deep_gemm_experts is None
and not envs.VLLM_ALLOW_BATCHED_TRITON_FALLBACK
):
raise RuntimeError(
"DeepGemm was requested but is not available. "
"The batched triton kernel fallback is disabled by default. "
"Set VLLM_ALLOW_BATCHED_TRITON_FALLBACK=1 to enable the fallback "
"for debugging purposes."
)
self.batched_triton_experts = (
BatchedTritonExperts(