Skip to content

spec-decode: honor the speculative-config attention backend for the DFlash draft #19

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
lukealonso merged 1 commit into from
Jun 14, 2026
Merged

spec-decode: honor the speculative-config attention backend for the DFlash draft #19

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
8 changes: 7 additions & 1 deletion vllm/v1/worker/gpu/spec_decode/dflash/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,13 @@ def load_dflash_model(target_model: nn.Module, vllm_config: VllmConfig) -> nn.Mo
draft_vllm_config = replace(
vllm_config,
attention_config=replace(
vllm_config.attention_config, use_non_causal=not causal
vllm_config.attention_config,
use_non_causal=not causal,
# Honor the speculative-config attention backend for the draft
# (matches llm_base_proposer): otherwise auto-select picks
# FlashInfer, which downgrades the spec-decode cudagraph to
# PIECEWISE and cannot do non-causal prefill under DCP.
backend=speculative_config.attention_backend,
),
)
with set_model_tag("dflash_head"):
Expand Down
Loading