Skip to content

[perf] slightly imporve DeepSeek-R1-FP4 TP8 #7481

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 1 commit into from
Jun 23, 2025
Merged

[perf] slightly imporve DeepSeek-R1-FP4 TP8 #7481

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
6 changes: 4 additions & 2 deletions python/sglang/srt/models/deepseek_v2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -362,12 +362,14 @@ def forward(
return self.forward_deepep(hidden_states, forward_batch)

def forward_normal_dual_stream(self, hidden_states: torch.Tensor) -> torch.Tensor:
# router_logits: (num_tokens, n_experts)
router_logits = self.gate(hidden_states)
Comment on lines +365 to +366
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Jun 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

Moving the router logits calculation before the current_stream capture and alt_stream wait might introduce a slight overhead if the gate operation is not fully asynchronous. Consider profiling to ensure this change doesn't negatively impact performance, as it could potentially delay the start of the shared experts computation.


current_stream = torch.cuda.current_stream()
self.alt_stream.wait_stream(current_stream)
shared_output = self._forward_shared_experts(hidden_states)

with torch.cuda.stream(self.alt_stream):
# router_logits: (num_tokens, n_experts)
router_logits = self.gate(hidden_states)
final_hidden_states = self.experts(
hidden_states=hidden_states, router_logits=router_logits
)
Expand Down
Loading