Add Audio Spectogram Transformer #19981
NielsRogge
merged
37 commits into
huggingface:main
from
NielsRogge:add_ast
First draft
8c104642
Make conversion script work
353ead3b
Add id2label mapping, run code quality
f996abe0
Fix copies
7006ed88
Add first draft of feature extractor
488e7e54
Update conversion script to use feature extractor
6a27c2ee
Make more tests pass
e8feefc9
Add docs
c0ec2685
update input_features to input_values + pad by default to max length
d952e5f6
Fix doc tests
3547d7c0
Add feature extractor tests
00471227
Add proper padding/truncation to feature extractor
56aafc53
Add support for conversion of all audioset checkpoints
f194e9d8
Improve docs and extend conversion script
df8575c3
Fix README
a08338f0
Rename spectogram to spectrogram
16285362
Fix copies
9dede0c3
Add integration test
fdead748
Remove dummy conv
70c948a5
Update to ast
fd10b76d
Update organization
341ade2e
Fix init
5685d730
Rename model to AST
a0e7d506
Add require_torchaudio annotator
19cf9f6b
Move import of ASTFeatureExtractor under a is_speech_available
6588ebaa
Fix rebase
3de16865
Add pipeline config
3b367978
Update name of classifier head
0a8afcae
Rename time_dimension and frequency_dimension for clarity
4deea23b
Remove print statement
8113ed12
Fix pipeline test
3154869b
Fix pipeline test
4282f68e
Fix index table
7e38d962
Fix init
519481cf
Fix conversion script
5dbf8998
Rename to ForAudioClassification
369ed798
Fix index table
b8e31b92
Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub