Skip to content

fix gibberish when use flaggems atten #382

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
cyber-pioneer merged 2 commits into from
Aug 19, 2026
Merged

fix gibberish when use flaggems atten #382

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4969838
fix Gibberish when use flaggems atten
ZenithJoyH Aug 17, 2026
57f0ff8
Merge branch 'flagos-ai:main' into main
ZenithJoyH Aug 18, 2026
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
5 changes: 5 additions & 0 deletions vllm_fl/dispatch/backends/flaggems/impl/attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,11 @@

class AttentionFLBackend(AttentionBackend):
accept_output_buffer: bool = True
# KV cache update is performed by vLLM via do_kv_cache_update() before
# forward(). Without this, vLLM assumes the backend updates the KV cache
# inside forward() and never calls do_kv_cache_update, leaving the cache
# unwritten (all zeros) and producing garbage output.
forward_includes_kv_cache_update: bool = False
supported_dtypes: ClassVar[list[torch.dtype]] = [torch.float16, torch.bfloat16]

@staticmethod
Expand Down
Loading