Skip to content

[None][fix] Fix KV cache recompute in draft_target spec decode #7348

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
mikeiovine merged 2 commits into from
Sep 3, 2025
Merged

[None][fix] Fix KV cache recompute in draft_target spec decode #7348

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
43ff0e8
[None][fix] Fix KV cache recompute in draft_target spec decode
mikeiovine Aug 28, 2025
20ef899
Merge branch 'main' into fix-recompute-kv
mikeiovine Sep 2, 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: 4 additions & 2 deletions tensorrt_llm/_torch/speculative/model_drafter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,8 +151,10 @@ def _create_draft_request_for_request(
assert num_draft_tokens == 0
return self._create_context_request(request, input_tokens)

# No tokens accepted - generation request
elif num_accepted_tokens == 0:
# No tokens accepted - generation request. This only applies to speculation algorithms
# that need to recompute KV cache for accepted tokens like eagle3.
elif num_accepted_tokens == 0 or not self.spec_config.spec_dec_mode.needs_kv_cache_recompute(
):
return self._create_generation_request(request, input_tokens)

# Tokens accepted - chunked context request
Expand Down