Skip to content

Fix triton kernels topk with keyword arguments #8732

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
Aug 3, 2025
Merged

Fix triton kernels topk with keyword arguments #8732

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a0c423c
update
ispobock Aug 3, 2025
96f16d4
Update python/sglang/srt/layers/moe/topk.py
ispobock Aug 3, 2025
b34a2c2
update
ispobock Aug 3, 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
24 changes: 21 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 @@ -183,12 +183,15 @@ def forward_cuda(
*,
num_token_non_padded: Optional[torch.Tensor] = None,
expert_location_dispatch_info: Optional[ExpertLocationDispatchInfo] = None,
sm_first: bool = False, # only used for triton kernels topk
) -> TopKOutput:
if self.use_triton_kernels:
routing_data, gather_idx, scatter_idx = routing(
router_logits, self.top_k, self.renormalize
return triton_kernels_topk(
router_logits=router_logits,
topk=self.top_k,
renormalize=self.renormalize,
sm_first=sm_first,
)
return TritonKernelTopKOutput(routing_data, gather_idx, scatter_idx)
else:
torch_native = False
return select_experts(
Expand Down Expand Up @@ -644,6 +647,21 @@ def biased_grouped_topk_cpu(
)


def triton_kernels_topk(
router_logits: torch.Tensor,
topk: torch.Tensor,
renormalize: bool = False,
sm_first: bool = False,
):
assert renormalize == False, "Triton kernels topk doesn't support renormalize"
routing_data, gather_idx, scatter_idx = routing(
logits=router_logits,
n_expts_act=topk,
sm_first=sm_first,
)
return TritonKernelTopKOutput(routing_data, gather_idx, scatter_idx)


if _is_cpu and _is_cpu_amx_available:
biased_grouped_topk = biased_grouped_topk_cpu
grouped_topk = grouped_topk_cpu
Expand Down
Loading