Skip to content

asr: bad audio fails the request instead of unloading the model - #190

Merged
SearchSavior merged 2 commits into
SearchSavior:mainfrom
meatposes:audio-bad-input
Oct 6, 2026
Merged

SearchSavior merged 2 commits into
SearchSavior:mainfrom
meatposes:audio-bad-input

Conversation

@meatposes

Copy link
Copy Markdown
Contributor

send whisper or qwen3-asr a file it can't decode (a webm from a browser, an m4a, whatever) and it throws inside the worker, which counts as an inference failure and unloads the model. one bad upload takes it down until someone reloads it.

decode errors now raise AudioDecodeError, which is a ValueError so the route returns 400, and the worker fails just that request. model stays loaded.

B50, whisper + qwen3-asr loaded, then a webm, an m4a and some junk bytes:

response models loaded after
main 500, then nothing works none
this PR 400 both

meatposes and others added 2 commits October 3, 2026 20:56
uploading something whisper or qwen3-asr can't decode raised inside the
worker, which marks it an inference failure and unloads the model. one
garbage upload took the model down until someone reloaded it.

decode errors now raise AudioDecodeError (a ValueError, so the route returns
400) and the worker fails just that request via packet.request_error. the
model stays loaded.
@SearchSavior
SearchSavior merged commit 9b0f0a5 into SearchSavior:main Oct 6, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants