Skip to content

[perf] Use CPU tensor to reduce GPU->CPU sync #25884

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
Sep 30, 2025
Merged

[perf] Use CPU tensor to reduce GPU->CPU sync #25884

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
9a6d1fc
[perf] Use CPU tensor to reduce GPU->CPU sync
lhtin Sep 29, 2025
0617ab6
[perf] Use .max_seq_len replace .seq_lens_cpu.max()
lhtin Sep 29, 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
2 changes: 1 addition & 1 deletion vllm/v1/worker/gpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2467,7 +2467,7 @@ def propose_draft_token_ids(sampled_token_ids):
effective_drafter_max_model_len = (
self.speculative_config.draft_model_config.max_model_len)
input_fits_in_drafter = spec_decode_common_attn_metadata and (
spec_decode_common_attn_metadata.seq_lens.max() +
spec_decode_common_attn_metadata.max_seq_len +
self.speculative_config.num_speculative_tokens
<= effective_drafter_max_model_len)
if use_padded_batch_for_eagle and input_fits_in_drafter:
Expand Down