Skip to content

Commit d86e3f0

Browse files
committed
lint
Signed-off-by: Bill Nell <bnell@redhat.com>
1 parent 3ca8322 commit d86e3f0

File tree

1 file changed

+4
-7
lines changed

1 file changed

+4
-7
lines changed

vllm/model_executor/layers/fused_moe/fused_batched_moe.py

Lines changed: 4 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -614,13 +614,10 @@ def apply(
614614
else:
615615
num = int(expert_num_tokens[expert].item())
616616
tmp = _resize_cache(workspace2, (num, N))
617-
if self.use_fp8_w8a8:
618-
assert False # TBD
619-
else:
620-
input = hidden_states[expert, :num, :] @ w1[expert].transpose(
621-
0, 1)
622-
self.activation(activation, tmp, input)
623-
out[expert, :num, :] = tmp @ w2[expert].transpose(0, 1)
617+
assert not self.use_fp8_w8a8
618+
input = hidden_states[expert, :num, :] @ w1[expert].transpose(0, 1)
619+
self.activation(activation, tmp, input)
620+
out[expert, :num, :] = tmp @ w2[expert].transpose(0, 1)
624621

625622
return out
626623

0 commit comments

Comments
 (0)