Skip to content

Remove DeepGEMM for indexer GEMM in piecewise NSA path #26494

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
May 28, 2026
Merged

Remove DeepGEMM for indexer GEMM in piecewise NSA path #26494

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
7 changes: 1 addition & 6 deletions python/sglang/srt/layers/attention/dsa/dsa_indexer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -162,12 +162,7 @@ def logits_head_gate_pcg(
softmax_scale: float,
q_scale: torch.Tensor,
) -> torch.Tensor:
from sglang.srt.layers.deep_gemm_wrapper import entrypoint as deep_gemm_wrapper

out = torch.empty(
(x.shape[0], weight.shape[0]), dtype=torch.float32, device=x.device
)
deep_gemm_wrapper.gemm_nt_bf16bf16f32(x, weight, out)
out = torch.mm(x, weight.t(), out_dtype=torch.float32)
Comment thread
Fridge003 marked this conversation as resolved.
weights = out * n_heads_inv_sqrt
weights = weights.unsqueeze(-1) * q_scale * softmax_scale
return weights
Expand Down
Loading