GitLab

Summary:
Previous Issue: --use-tmp-hub-dir expected the temp directories used to store large file to be deleted after each test case, but pytest erases directories after 3 full test sessions. This commit fixes by manually deleting a new subdirectory created in each test case. https://github.com/pytorch/audio/pull/2565#discussion_r929007101

Pull Request resolved: https://github.com/pytorch/audio/pull/2569

Reviewed By: nateanl

Differential Revision: D38117848

Pulled By: skim0514

fbshipit-source-id: 3767cb8df1238fd6218f6aaa58d5d583cea72699

Summary:
- Add documentation page for `SourceSeparationBundle` and `CONVTASNET_BASE_LIBRI2MIX`.
- Add citation of Libri2Mix dataset in the bundle documentation.
- url in integration test should use slash instead of `os.path.join` as it will fail on Windows. Change it to f-string.

Pull Request resolved: https://github.com/pytorch/audio/pull/2559

Reviewed By: carolineechen

Differential Revision: D38036116

Pulled By: nateanl

fbshipit-source-id: 736732805191113955badfec3955e2e24e8f4836

Summary:
- Add SourceSeparationBundle class for source separation pipeline
- Add `CONVTASNET_BASE_LIBRI2MIX` that is trained on Libri2Mix dataset.
- Add integration test with example mixture audio and expected scale-invariant signal-to-distortion ratio (Si-SDR) score. The test computes the Si-SDR score with permutation-invariant training (PIT) criterion for all permutations of sources and use the highest value as the final output. The test verifies if the score is equal to or larger than the expected value.

Pull Request resolved: https://github.com/pytorch/audio/pull/2440

Reviewed By: mthrok

Differential Revision: D37997646

Pulled By: nateanl

fbshipit-source-id: c951bcbbe8b7ed9553cb8793d6dc1ef90d5a29fe

Summary:
In https://github.com/pytorch/audio/issues/2283, torchaudio's downloading function is updated to reduce code duplication. The links in `EMFORMER_RNNT_BASE_LIBRISPEECH` are updated, but the ones in prototype pipelines are not. This PR addresses it by updating the download links of `EMFORMER_RNNT_BASE_MUSTC` and `EMFORMER_RNNT_BASE_TEDLIUM3` in prototype. Corresponding integration tests are added as well.

Pull Request resolved: https://github.com/pytorch/audio/pull/2444

Reviewed By: mthrok

Differential Revision: D37389178

Pulled By: nateanl

fbshipit-source-id: 46598dd71c95be47d1e1b54cef89ea51d280e17a

Summary:
Move CTC beam search decoder out of prototype to new `torchaudio.models.decoder` module.

hwangjeff mthrok any thoughts on the new module + naming, and if we should move rnnt beam search here as well??

Pull Request resolved: https://github.com/pytorch/audio/pull/2410

Reviewed By: mthrok

Differential Revision: D36784521

Pulled By: carolineechen

fbshipit-source-id: a2ec52f86bba66e03327a9af0c5df8bbefcd67ed

Summary:
Applies new import merging and sorting from µsort v1.0.

When merging imports, µsort will make a best-effort to move associated
comments to match merged elements, but there are known limitations due to
the diynamic nature of Python and developer tooling. These changes should
not produce any dangerous runtime changes, but may require touch-ups to
satisfy linters and other tooling.

Note that µsort uses case-insensitive, lexicographical sorting, which
results in a different ordering compared to isort. This provides a more
consistent sorting order, matching the case-insensitive order used when
sorting import statements by module name, and ensures that "frog", "FROG",
and "Frog" always sort next to each other.

For details on µsort's sorting and merging semantics, see the user guide:
https://usort.readthedocs.io/en/stable/guide.html#sorting

Reviewed By: lisroach

Differential Revision: D36402214

fbshipit-source-id: b641bfa9d46242188524d4ae2c44998922a62b4c

Summary:
Add support for lexicon free decoding based on [fairseq's](https://github.com/pytorch/fairseq/blob/main/examples/speech_recognition/new/decoders/flashlight_decoder.py#L53) implementation. Reached numerical parity with fairseq's decoder in offline experimentation

Follow ups
- Add pretrained LM support for lex free decoding
- Add example in tutorial
- Replace flashlight C++ source code with flashlight text submodule
- [optional] fairseq compatibility test

Pull Request resolved: https://github.com/pytorch/audio/pull/2342

Reviewed By: nateanl

Differential Revision: D35856104

Pulled By: carolineechen

fbshipit-source-id: b64286550984df906ebb747e82f6fb1f21948ac7

Summary:
PyTorch Lite, which is becoming a standard for mobile PyTorch usage, does not support containers containing custom classes. Consequently, because TorchAudio's RNN-T decoder currently returns and accepts lists of `Hypothesis` namedtuples, it is not compatible with PyTorch Lite. This PR resolves said incompatibility by changing the underlying implementation of `Hypothesis` to tuple.

Pull Request resolved: https://github.com/pytorch/audio/pull/2339

Reviewed By: nateanl

Differential Revision: D35806529

Pulled By: hwangjeff

fbshipit-source-id: 9cbae5504722390511d35e7f9966af2519ccede5

Summary:
add function to download pretrained files for LibriSpeech 3-gram/4-gram KenLM, tests, and updated tutorial

Pull Request resolved: https://github.com/pytorch/audio/pull/2275

Reviewed By: mthrok

Differential Revision: D35115418

Pulled By: carolineechen

fbshipit-source-id: 83ff22380fce9c753bb4a7b7e3d89aa66c2831c0

Summary:
In recent updates, torchaudio added features that download assets/models from
download.pytorch.org/torchaudio.

To reduce the code duplication, the implementations uses utilities from
``torch.hub``, but still, there are patterns repeated in implementing
the fetch mechanism, notably cache and local file path handling.

This commit introduces the utility function that handles
download/cache/local path management that can be used for
fetching pre-trained model data.

Pull Request resolved: https://github.com/pytorch/audio/pull/2283

Reviewed By: carolineechen

Differential Revision: D35050469

Pulled By: mthrok

fbshipit-source-id: 219dd806f9a96c54d2d31e981c1bbe282772702b

Summary:
Moves ASR features out of `torchaudio.prototype`. Specifically, merges contents of `torchaudio.prototype.models` into `torchaudio.models` and contents of `torchaudio.prototype.pipelines` into `torchaudio.pipelines` and updates refs, tests, and docs accordingly.

Pull Request resolved: https://github.com/pytorch/audio/pull/2187

Reviewed By: nateanl, mthrok

Differential Revision: D33918092

Pulled By: hwangjeff

fbshipit-source-id: f003f289a7e5d7d43f85b7c270b58bdf2ed6344c

Summary:
Adds integration test for pretrained ASR pipeline `EMFORMER_RNNT_BASE_LIBRISPEECH`.

Pull Request resolved: https://github.com/pytorch/audio/pull/2172

Reviewed By: carolineechen, nateanl

Differential Revision: D33793324

Pulled By: hwangjeff

fbshipit-source-id: d0613e2ab98fe5afa7b16ca39b67f0a0304d13fc

Summary:
cc mthrok

Pull Request resolved: https://github.com/pytorch/audio/pull/2116

Reviewed By: mthrok

Differential Revision: D33368453

Pulled By: jdsgomes

fbshipit-source-id: 09cf3fe5ed6f771c2f16505633c0e59b0c27453c

Summary:
Pull Request resolved: https://github.com/pytorch/audio/pull/2096

run: `arc lint --apply-patches --paths-cmd 'hg files -I "./**/*.py"'`

Reviewed By: mthrok

Differential Revision: D33297351

fbshipit-source-id: 7bf5956edf0717c5ca90219f72414ff4eeaf5aa8

This commit changes all the `torch.hub` network utility functions to
be imported from `torchaudio._internal`, so that later we can replace
the function within fbcode.

* Add wav2vec2 ASR German pretrained model from voxpopuli

- Make the test support other languages
- Fetch tetst asset on-the-fly

* [BC-breaking] Remove unused dimension from pretrained Wav2Vec2 ASR

The Wav2Vec2 ASR pretrained weights originated from fairseq have
extra dimension that have nothing to do with the ASR task.

https://github.com/pytorch/fairseq/blob/c5ff181125c7e6126b49a85e5ebdd5f5b6a07914/fairseq/data/dictionary.py#L18-L37

which is masked during the loss computation as

https://github.com/pytorch/fairseq/blob/c5ff181125c7e6126b49a85e5ebdd5f5b6a07914/fairseq/criterions/ctc.py#L126-L128

This change removes it.

* Use '-' for blank token representation.

Future work items:
- length computation of GriffinLim
- better way to make InverseMelScale work in inference_mode

- Move wav2vec2 pretrained weights to `torchaudio.pipelines` namespace to align with #1872.
- Split `Wav2Vec2PretrainedModelBundle` into `Wav2Vec2Bundle` (for pre-training model) and  `Wav2Vec2ASRBundle` (for models fine-tuned for ASR).
- Update base URL

Add pretrained weights from https://github.com/pytorch/fairseq/tree/main/examples/wav2vec#pre-trained-models
- Wav2Vec 2.0 Base / Large / Large (LV-60)
- XLSR-53

This commit adds
- HUBERT_LARGE
- HUBERT_XLARGE
- HUBERT_ASR_XLARGE