[ASR] Add official ASR CTC example to examples/pytorch/speech-recognition#13620
Merged
patrickvonplaten merged 33 commits intoSep 24, 2021
Merged
Conversation
This was referenced Sep 17, 2021
…into add_asr_example
…into add_asr_example
| # 3. Next, we create the vocabulary of the model by extracting all unique characters from | ||
| # the training and evaluation datasets | ||
| # We need to make sure that only first rank saves vocabulary | ||
| if training_args.world_size == 1 or dist.get_rank() == 0: |
Contributor
Author
There was a problem hiding this comment.
this caused me an headache for 3 days -> in distributed training each process was creating a different ordering of characters in the vocabulary which essentially meant that each process has different label ids.
By using sorted(...) and making sure that only the first process creates & saves the vocabulary, the problem is solved.
patrickvonplaten
changed the title
examples/pytorch/speech-recognitionexamples/pytorch/speech-recognition
patrickvonplaten
requested review from
albertvillanova,
anton-l,
lhoestq,
patil-suraj and
sgugger
sgugger
approved these changes
Sep 22, 2021
sgugger
left a comment
sgugger
left a comment
Collaborator
There was a problem hiding this comment.
Thanks a lot for adding this example and great job figuring out the problem in a distributed setup!
patil-suraj
approved these changes
Sep 23, 2021
patil-suraj
left a comment
patil-suraj
left a comment
Contributor
There was a problem hiding this comment.
Looks really good! Thanks for adding this example
patil-suraj
reviewed
Sep 23, 2021
Co-authored-by: Suraj Patil <surajp815@gmail.com>
anton-l
approved these changes
Sep 23, 2021
anton-l
left a comment
anton-l
left a comment
Member
There was a problem hiding this comment.
Looks good, thank you very much for figuring out the DDP problem!
The torchaudio loader seems to be the best fit for the example 🙂
Although I think Windows users will be out of luck when they try to load mp3's (soundfile is used as a backend there, and it specifically excludes mp3: http://www.mega-nerd.com/libsndfile/#Features)
P.S. So sorry for the typo spam 😅
Co-authored-by: Anton Lozhkov <aglozhkov@gmail.com>
Co-authored-by: Anton Lozhkov <aglozhkov@gmail.com>
…/transformers into add_asr_example
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
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.
This PR adds a generic speech recognition for CTC example. It has been tested for single GPU and distributed training on Common Voice and is being tested on Librispeech currently.
Once
datasetshas https://github.com/huggingface/datasets/pull/2324/files merged and made a new release I will slightly adapt the script to leverage the new audio feature.A couple of example runs with this script:
This example folder should have two additional scripts: 1 for Seq2Seq ASR + 1 for CTC + LM decoding which are left for future work