Skip to content

Commit 2212cd6

Browse files
varun-sundar-rabindranathVarun Sundar Rabindranath
andauthored
[Bugfix] DeepGemm utils : Fix hardcoded type-cast (vllm-project#21517)
Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com> Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
1 parent ce3a9b1 commit 2212cd6

File tree

1 file changed

+1
-1
lines changed

1 file changed

+1
-1
lines changed

vllm/model_executor/layers/fused_moe/deep_gemm_utils.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -52,7 +52,7 @@ def compute_aligned_M(M: int, num_topk: int, local_num_experts: int,
5252
@triton.jit
5353
def apply_expert_map(expert_id, expert_map):
5454
if expert_id != -1:
55-
expert_id = tl.load(expert_map + expert_id).to(tl.int64)
55+
expert_id = tl.load(expert_map + expert_id).to(expert_id.dtype)
5656
return expert_id
5757

5858

0 commit comments

Comments
 (0)