Word-level timestamps broken for short-form audio #30325
amyeroberts
merged
10 commits into
huggingface:main
from
kamilakesbi:whisper_large_fix_word_level_timestamps
force chunk_length_s in AutomaticSpeechRecognitionPipeline
fecd79d3
compute num_frames even when stride is None
0f5fbae9
add slow tests
24a340a9
sanchit-gandhi
approved these changes
on 2024-04-23
fix test
15395df9
Update src/transformers/pipelines/automatic_speech_recognition.py
034449cc
Update tests/pipelines/test_pipelines_automatic_speech_recognition.py
e52d11a5
kamilakesbi
changed the title
add input validation
47da6410
Merge branch 'whisper_large_fix_word_level_timestamps' of github.com:…
a8cd4832
fixup
76fbfd35
amyeroberts
approved these changes
on 2024-05-02
small fix
a7a5d5e6
Login to write a write a comment.
Login via GitHub