Skip to content

feat(low_latency): support top-16 routing and hidden-dim 3584 (Kimi-K3) #11

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
Fridge003 merged 1 commit into from
Aug 19, 2026
Merged

feat(low_latency): support top-16 routing and hidden-dim 3584 (Kimi-K3) #11

Show file tree
Hide file tree
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
4 changes: 2 additions & 2 deletions csrc/kernels/legacy/internode_ll.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -702,7 +702,7 @@ void dispatch(void* packed_recv_x,
int num_device_sms,
cudaStream_t stream,
int phases) {
constexpr int kNumMaxTopK = 11;
constexpr int kNumMaxTopK = 16; // Kimi-K3 routes top-16
const int num_warp_groups = ceil_div(num_experts, num_device_sms);
const int num_warps_per_group = 32 / num_warp_groups;
EP_HOST_ASSERT(num_warp_groups > 0 and num_warps_per_group > 0);
Expand Down Expand Up @@ -1880,7 +1880,7 @@ void combine(void* combined_x,
);
}

constexpr int kNumMaxTopk = 11;
constexpr int kNumMaxTopk = 16; // Kimi-K3 routes top-16
const int num_warp_groups = ceil_div(num_experts, num_device_sms);
const int num_warps_per_group = 32 / num_warp_groups;
const int num_recv_per_sm = ceil_div(num_combined_tokens, num_device_sms);
Expand Down
2 changes: 2 additions & 0 deletions csrc/kernels/legacy/launch.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,6 +118,8 @@
case_macro(2560); \
case 3072: \
case_macro(3072); /* for gpt-oss */ \
case 3584: \
case_macro(3584); /* for kimi-k3 */ \
case 4096: \
case_macro(4096); \
case 5120: \
Expand Down