Torchaudio fixes #2007
Torchaudio fixes #2007
Conversation
| @@ -220,7 +223,7 @@ def _transform_to_feature( | |||
| ): | |||
| feature_type = audio_feature_dict[TYPE] | |||
| if feature_type == "raw": | |||
| audio_feature = np.expand_dims(audio, axis=-1) | |||
| audio_feature = np.expand_dims(audio[0], axis=-1) | |||
There was a problem hiding this comment.
Is there a regression test we can add for the hotfix?
There was a problem hiding this comment.
Ya good point. So since audio is a less tested feature right now, we definitely need better testing for it. Getting a test together for the the "raw" feature type in addition to a test for the remote capabilities is another (since I've been testing with an internal s3 bucket which seems a bit unorthodox). Since I need to have an audio model trained by Wednesday for the KP presentation however, I kinda need to move quick to get that together before I build out a test suite for the rest of the audio feature pipeline. I should be able to make a lot of progress through the week though assuming the default changes allow me to create the in Predibase.
There was a problem hiding this comment.
Got it, sounds good!
* first commit * added regression test * cleanup * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add ray tests * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Co-authored-by: Geoffrey Angus <geoffrey@predibase.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This fix should fix the issues we were seeing in the audio models