Add Speech Seq2Seq Training script #14792
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Hmmm, not that easy the training it seems...will run some more examples next week with Wav2Vec2 - BERT |
|
Getting good results now for wav2vec2 - bart: https://huggingface.co/patrickvonplaten/wav2vec2-2-bart-base |
|
Results are good: https://huggingface.co/models?other=asr_seq2esq Cleaning up the PR and we can merge in 1,2 days |
patrickvonplaten
changed the title
sgugger
approved these changes
Dec 27, 2021
There was a problem hiding this comment.
Nice new example! Two comments before merging:
- the test is the second longest at 47s. If it's possible to reduce it a little bit, that would be great.
- We already had a discussion on the fact that the speech models have a
feature_extractorattribute that is distinct from the feature extractor that preprocesses the data, which is very confusing. We need another name for thefreeze_feature_extractormethods and argument as a result.
examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py
Outdated
Show resolved
Hide resolved
examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py
Outdated
Show resolved
Hide resolved
examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py
Outdated
Show resolved
Hide resolved
examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py
Outdated
Show resolved
Hide resolved
src/transformers/models/speech_encoder_decoder/modeling_speech_encoder_decoder.py
Show resolved
Hide resolved
|
Merging this now. Very much agree on the naming issue @sgugger and thanks for reminding me again. Will open another PR for this later today. |
stevhliu
pushed a commit
to stevhliu/transformers
that referenced
this pull request
Jan 6, 2022
* start * add gradient checkpointing and feature extractor freezing * Apply suggestions from code review * up * up * up * correct * up * more changes * up * up * up * remove rst
Albertobegue
pushed a commit
to Albertobegue/transformers
that referenced
this pull request
Jan 27, 2022
* start * add gradient checkpointing and feature extractor freezing * Apply suggestions from code review * up * up * up * correct * up * more changes * up * up * up * remove rst
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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?
An explanation of this new training script is given on the README.md.
Two successful training runs can be seen here: https://huggingface.co/models?other=asr_seq2esq
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.