As noted in an other PR by @QuentinFuxa, it seems all recent PRs fail on the same flaky or faulty test:
FAILED test_transformers.py::TestWav2Vec2::test_transformers_wav2vec2[facebook/wav2vec2-large-robust-ft-swbd-300h-expected_transcription0-cpu]
Expected: MISTER QUILTER IS THE APOSSEL OF THE MIDDLE CLASSES AND WE ARE GLAD TO WELCOME HIS GOSPEL
Got: MISTER QUILTER IS THE APOSSTEL OF THE MIDDLE CLASSES AND WE ARE GLAD TO WELCOME HIS GOSPEL
The model produces a slightly different transcription (APOSSTEL vs APOSSEL) likely due to environment/library version differences. All other 166 tests passed. Could a maintainer re-run the failed job?
Originally posted by @QuentinFuxa in #2010 (comment)
As noted in an other PR by @QuentinFuxa, it seems all recent PRs fail on the same flaky or faulty test:
The model produces a slightly different transcription (
APOSSTELvsAPOSSEL) likely due to environment/library version differences. All other 166 tests passed. Could a maintainer re-run the failed job?Originally posted by @QuentinFuxa in #2010 (comment)