Skip to content

fused_moe_kernel - cast accumulator after applying router weights #32002

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
DarkLight1337 merged 1 commit into from
Jan 10, 2026
+5 −8
Merged

fused_moe_kernel - cast accumulator after applying router weights #32002

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 5 additions & 8 deletions vllm/model_executor/layers/fused_moe/fused_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -520,14 +520,9 @@ def fused_moe_kernel(
b_ptrs += BLOCK_SIZE_K * stride_bk

if use_int8_w8a16:
accumulator = (accumulator * b_scale).to(compute_type)
elif use_fp8_w8a8 or use_int8_w8a8:
if group_k > 0 and group_n > 0:
accumulator = accumulator.to(compute_type)
else:
accumulator = (accumulator * a_scale * b_scale).to(compute_type)
else:
accumulator = accumulator.to(compute_type)
accumulator = accumulator * b_scale
elif (use_fp8_w8a8 or use_int8_w8a8) and not (group_k > 0 and group_n > 0):
accumulator = accumulator * a_scale * b_scale

# Since bias is typically not quantized, it's added after dequantization.
if HAS_BIAS:
Expand All @@ -536,6 +531,8 @@ def fused_moe_kernel(
moe_weight = tl.load(topk_weights_ptr + offs_token, mask=token_mask, other=0)
accumulator = accumulator * moe_weight[:, None]

accumulator = accumulator.to(compute_type)

# -----------------------------------------------------------
# Write back the block of the output
offs_cn = pid_n * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N)
Expand Down