Skip to content
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

[Speculative Decoding] Move indices to device before filtering output #10850

Merged
merged 1 commit into from
Dec 3, 2024
Merged

[Speculative Decoding] Move indices to device before filtering output #10850

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
9 changes: 6 additions & 3 deletions vllm/spec_decode/multi_step_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -120,6 +120,9 @@ def sampler_output(
indices_of_seq_with_bonus_tokens)
model_outputs.append(model_output)

# move indices to device to avoid stream sync
indices_of_seq_with_bonus_tokens = torch.tensor(
indices_of_seq_with_bonus_tokens, device=self.device)
filtered_model_outputs = self._filter_model_output(
model_outputs, indices_of_seq_with_bonus_tokens)
return filtered_model_outputs, True
Expand Down Expand Up @@ -189,7 +192,7 @@ def _expand_execute_model_request(
@staticmethod
def _filter_model_output(
expanded_batch_outputs: List[SamplerOutput],
output_indices_to_retain: List[int]) -> List[SamplerOutput]:
output_indices_to_retain: torch.Tensor) -> List[SamplerOutput]:
"""
Filters the model output to include only the specified sequence
outputs. This method contracts the expanded batch output from the
Expand All @@ -199,8 +202,8 @@ def _filter_model_output(
Args:
expanded_batch_output (List[SamplerOutput]): The expanded output
batch from the model.
output_indices_to_retain (List[int]): Indices of the model outputs
to retain.
output_indices_to_retain (torch.Tensor): Indices of the model
outputs to retain.

Returns:
List[SamplerOutput]: A list containing the filtered model
Expand Down
Loading