Replace input_values_processing with unpack_inputs #21502
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Updates Hubert and Wav2Vec2 to use the repo standard
unpack_inputsdecorator instead ofinput_values_processing. Previously, TensorFlow models useinput_values_processing, a function which processed inputs ready to be passed to the TF layers. This was removed and replaced with unpack inputs, see #16051. For some reason, wav2vec2 wasn't part of this update. A bug was reported, but the conversation and details can't be found in the discord 😢This PR also updates any relevant tests:
unittest.skipdecorator so tests are marked as skipped rather than passedNote: The loss calculation for
TFHubertForCTCwasn't being calculated properly. Without theunpack_inputsdecorator, whenmodel.fitwas being called,labelswere passed into the model as part of theinput_valuesdictionary. These were then processed and part of theinputsdictionary. However,labelswas always taken from the function kwargs and so loss calculation skipped. This is why the previous test 'returned the wrong shape'. The issue was resolved forTFWav2Vec2ForCTCin a previous PR #18014Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.