Skip to content

Fix Qwen3MoE missing token padding optimization #6820

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
merged 3 commits into from
Jun 5, 2025
Merged

Fix Qwen3MoE missing token padding optimization #6820

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
8248b87
more
fzyzcjy Jun 3, 2025
95772ca
more
fzyzcjy Jun 3, 2025
7c10795
Merge branch 'main' into feat/token_pad
yizhang2077 Jun 4, 2025
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
6 changes: 3 additions & 3 deletions python/sglang/srt/layers/moe/topk.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,6 +66,7 @@ def fused_topk(
gating_output: torch.Tensor,
topk: int,
renormalize: bool,
num_token_non_padded: Optional[torch.Tensor] = None,
expert_location_dispatch_info: Optional[ExpertLocationDispatchInfo] = None,
):
assert hidden_states.shape[0] == gating_output.shape[0], "Number of tokens mismatch"
Expand All @@ -91,6 +92,7 @@ def fused_topk(
if renormalize:
topk_weights = topk_weights / topk_weights.sum(dim=-1, keepdim=True)
topk_ids = topk_ids_logical_to_physical(topk_ids, expert_location_dispatch_info)
_mask_topk_ids_padded_region(topk_ids, num_token_non_padded)
return topk_weights, topk_ids


Expand Down Expand Up @@ -363,15 +365,13 @@ def select_experts(
renormalize=renormalize,
)
elif custom_routing_function is None:
assert (
num_token_non_padded is None
), "num_token_non_padded is not yet supported in fused_topk"
# Qwen3MOE uses fused_topk
topk_weights, topk_ids = fused_topk(
hidden_states=hidden_states,
gating_output=router_logits,
topk=top_k,
renormalize=renormalize,
num_token_non_padded=num_token_non_padded,
expert_location_dispatch_info=expert_location_dispatch_info,
)
else:
Expand Down
2 changes: 2 additions & 0 deletions python/sglang/srt/models/qwen3_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -193,6 +193,7 @@ def forward_deepep(
top_k=self.top_k,
use_grouped_topk=False,
renormalize=self.renormalize,
num_token_non_padded=forward_batch.num_token_non_padded,
expert_location_dispatch_info=ExpertLocationDispatchInfo.init_new(
layer_id=self.layer_id,
),
Expand Down Expand Up @@ -260,6 +261,7 @@ def op_select_experts(self, state):
top_k=self.top_k,
use_grouped_topk=False,
renormalize=self.renormalize,
num_token_non_padded=state.forward_batch.num_token_non_padded,
expert_location_dispatch_info=ExpertLocationDispatchInfo.init_new(
layer_id=self.layer_id,
),
Expand Down
Loading