Skip to content

Commit 484fc83

Browse files
committed
revert some bad changes
Signed-off-by: Bill Nell <bnell@redhat.com>
1 parent 40ebc47 commit 484fc83

File tree

2 files changed

+2
-1
lines changed

2 files changed

+2
-1
lines changed

vllm/model_executor/layers/fused_moe/layer.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -497,7 +497,7 @@ def forward_cuda(
497497
custom_routing_function=custom_routing_function,
498498
scoring_func=scoring_func,
499499
e_score_correction_bias=e_score_correction_bias,
500-
indices_type=torch.uint32 if self.use_pplx_kernels else None)
500+
indices_type=torch.uint32 if self.moe.use_pplx_kernels else None)
501501

502502
if self.rocm_aiter_moe_enabled:
503503
return self.rocm_aiter_fused_experts(

vllm/platforms/cuda.py

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -158,6 +158,7 @@ def check_and_update_config(cls, vllm_config: "VllmConfig") -> None:
158158
"currently not supported with CUDA Graphs.")
159159
vllm_config.model_config.enforce_eager = True
160160
compilation_config.use_cudagraph = False
161+
compilation_config.use_inductor = False
161162

162163
@classmethod
163164
def get_current_memory_usage(cls,

0 commit comments

Comments
 (0)