Skip to content

Fix dtype for idle input in spec decoding #7456

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 2 commits into from
Jun 23, 2025
Merged

Fix dtype for idle input in spec decoding #7456

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
842916c
fix dtype for idle input
ch-wan Jun 23, 2025
e7ecf67
format
ch-wan Jun 23, 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/speculative/eagle_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,14 +89,13 @@ def create_idle_input(
cls,
device: torch.device,
hidden_size: int,
dtype: torch.dtype,
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

Adding dtype here changes the order of arguments in the function signature. While this isn't inherently wrong, it's worth considering if it impacts readability or maintainability. Ensure that all call sites are updated accordingly. It might be better to keep the order consistent with the original function definition if there's no strong reason to change it.

topk: int,
capture_hidden_mode: CaptureHiddenMode,
):
return cls(
verified_id=None,
hidden_states=torch.empty(
(0, hidden_size), device=device, dtype=torch.float32
),
hidden_states=torch.empty((0, hidden_size), device=device, dtype=dtype),
topk_p=torch.empty((0, topk), device=device, dtype=torch.float32),
topk_index=torch.empty((0, topk), device=device, dtype=torch.int64),
capture_hidden_mode=capture_hidden_mode,
Expand Down Expand Up @@ -334,6 +333,7 @@ def verify(
draft_input=EagleDraftInput.create_idle_input(
device=batch.device,
hidden_size=batch.model_config.hidden_size,
dtype=batch.model_config.dtype,
topk=self.topk,
capture_hidden_mode=CaptureHiddenMode.LAST,
),
Expand Down
2 changes: 2 additions & 0 deletions python/sglang/srt/speculative/eagle_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -498,6 +498,7 @@ def _draft_preprocess_idle(self, batch: ScheduleBatch):
batch.spec_info = EagleDraftInput.create_idle_input(
device=self.device,
hidden_size=self.model_config.hidden_size,
dtype=self.model_config.dtype,
topk=self.topk,
capture_hidden_mode=CaptureHiddenMode.LAST,
)
Expand Down Expand Up @@ -838,6 +839,7 @@ def forward_draft_extend_after_decode(
batch.spec_info = EagleDraftInput.create_idle_input(
device=self.device,
hidden_size=self.model_config.hidden_size,
dtype=self.model_config.dtype,
topk=self.topk,
capture_hidden_mode=CaptureHiddenMode.LAST,
)
Expand Down
Loading