Skip to content

GitLab

Switch branch/tag
  1. 19 Sep, 2023 1 commit
  3. 13 Sep, 2023 1 commit
  5. 08 Sep, 2023 1 commit
  7. 04 Sep, 2023 2 commits
  9. 21 Aug, 2023 1 commit
  11. 20 Aug, 2023 1 commit
  13. 15 Aug, 2023 1 commit
    • moto's avatar
      [BC-breaking] Update pre-built ffmpeg4 to 4.4.4 (#3561) · bf07ea6b
      moto authored 
      Summary:
In https://github.com/pytorch/audio/pull/3460, we switched the build process for FFmpeg extension.
Since it is complicated to install FFmpeg in some environments, at build time, pre-built binaries and its headers
are downloaded and used as a scaffolding for torchaudio build.

Now even though we did not change any code or FFmpeg version, it turned out that this causes segmentation
fault on Ubuntu when using system Python and FFmpeg 4.4 installed via aptitude.
While investigating the issue, I swapped the said pre-built FFmpeg scaffolding with FFmpeg 4.4 from aptitude,
and the segmentation fault did not happen. This indicates that it is binary compatibility issue.

Before https://github.com/pytorch/audio/issues/3460, each binary build job was building FFmpeg 4.1.8 using the same compiler used to build torchaudio,
but after https://github.com/pytorch/audio/issues/3460 the environments to build FFmpeg 4.1.8 and torchaudio are different. My hypothesis is that
this difference is causing some ABI incompatibility when linking against FFmpeg 4.4. (Also, I don't remember well,
but I read somewhere that 4.4 has a different ABI)

Through experiments, it turned out upgrading the pre-built FFmpeg scaffolding to 4.4 resolves this.
So this commit upgrade the pre-built FFmpeg 4 to 4.4.
The potential (yet unconfirmed) downside is that torchaudio will no longer work with 4.1, 4.2, and 4.3.
Since FFmpeg 4.4 is what Ubuntu 20.04 and 22.04 support by default, and Google Colab is also on 20.04,
I think it is more important to support 4.4.

Therefore we drop the support for 4.1-4.3 from normal build (and official distributions). Those who wish to
use 4.1-4.3 can build torchaudio from source by linking to specific FFmpeg.

Pull Request resolved: https://github.com/pytorch/audio/pull/3561

Reviewed By: hwangjeff

Differential Revision: D48340201

Pulled By: mthrok

fbshipit-source-id: 7ece82910f290c7cf83f58311c4cf6a384e8795e
      bf07ea6b
  15. 10 Aug, 2023 1 commit
  17. 08 Aug, 2023 4 commits
  19. 04 Aug, 2023 1 commit
  21. 01 Aug, 2023 1 commit
  23. 31 Jul, 2023 2 commits
  25. 29 Jul, 2023 1 commit
    • moto's avatar
      Refactor compat (#3518) · 8497ee91
      moto authored 
      Summary:
The I/O functions in _compat module was introduced there so that
everything related to FFmpeg is in torchaudio.io and FFmpeg library
initialization can be carried out in `torchaudio.io.__init__`.

Now that this constraint is removed, (all the initialization happens
at `torchaudio._extension.__init__`) and `_compat` is only used by
FFmpeg dispatcher backend, we move the module to `torchaudio._backend`
for better locality.

Pull Request resolved: https://github.com/pytorch/audio/pull/3518

Reviewed By: huangruizhe

Differential Revision: D47877412

Pulled By: mthrok

fbshipit-source-id: aa18c8cb6e5d5360950df5158c33c653e37c565f
      8497ee91
  27. 28 Jul, 2023 2 commits
  29. 26 Jul, 2023 1 commit
  31. 25 Jul, 2023 2 commits
    • Pingchuan Ma's avatar
      Update avsr recipe (#3493) · d4644793
      Pingchuan Ma authored 
      Summary:
This PR is to include few changes in the AV-ASR recipe. The changes include better results, a faster face detector (Mediapipe), renamed variable names, a streamlined dataloader, and a few illustrated examples. These changes were made to improve the usability of the recipe.

Pull Request resolved: https://github.com/pytorch/audio/pull/3493

Reviewed By: mthrok

Differential Revision: D47758072

Pulled By: mpc001

fbshipit-source-id: 4533587776f3a7a74f3f11b0ece773a0934bacdc
      d4644793
    • moto's avatar
      Update nvdec/nvenc tutorials (#3483) · 56e22664
      moto authored 
      Summary: Pull Request resolved: https://github.com/pytorch/audio/pull/3483

Differential Revision: D47725664

Pulled By: mthrok

fbshipit-source-id: e4249e1488fa7af8670be4a5077957912ff3420b
      56e22664
  33. 24 Jul, 2023 1 commit
  35. 18 Jul, 2023 1 commit
  37. 15 Jul, 2023 1 commit
  39. 05 Jul, 2023 1 commit
  41. 28 Jun, 2023 1 commit
  43. 26 Jun, 2023 1 commit
  45. 21 Jun, 2023 1 commit
  47. 16 Jun, 2023 1 commit
    • Pingchuan Ma's avatar
      Add LRS3 data preparation (#3421) · 77cdd160
      Pingchuan Ma authored 
      Summary:
This PR adds a data preparation recipe that uses the ultra face detector to extract full-face video. The resulting video output is then used as input for training and evaluating RNNT-based models for automatic speech recognition (ASR), visual speech recognition (VSR), and audio-visual ASR (AV-ASR) on the LRS3 dataset.

This PR also updates the word error rate (WER) for AV-ASR LRS3 models and improves the code readability.

Pull Request resolved: https://github.com/pytorch/audio/pull/3421

Reviewed By: mpc001

Differential Revision: D46799748

Pulled By: mthrok

fbshipit-source-id: 97af3feac0592b240617faaffa4c0ac8cef614a9
      77cdd160
  49. 15 Jun, 2023 1 commit
    • moto's avatar
      Update forced alignment tutorial (#3440) · 18601691
      moto authored 
      Summary:
* Fix backtrack visualization (the cooridnate was off-by-one.)
* Add note about the simplification and the new align API
* Explicitly handle SOS and EOS

Pull Request resolved: https://github.com/pytorch/audio/pull/3440

Reviewed By: xiaohui-zhang

Differential Revision: D46761282

Pulled By: mthrok

fbshipit-source-id: b0b6c9754674e8e23543e9f002e29b55102c92f8
      18601691
  51. 07 Jun, 2023 1 commit
  53. 06 Jun, 2023 1 commit
  55. 04 Jun, 2023 1 commit
  57. 02 Jun, 2023 2 commits
    • moto's avatar
      [BC-Breaking] Remove compute_kaldi_pitch (#3368) · 5bbbb1d5
      moto authored 
      Summary:
This commit removes compute_kaldi_pitch function and the underlying Kaldi integration from torchaudio.

Kaldi pitch function was added in a short period of time by integrating the original Kaldi implementation, instead of reimplementing it in PyTorch.

The Kaldi integration employed a hack which replaces the base vector/matrix implementation of Kaldi with PyTorch Tensor so that there is only one blas library within torchaudio.

Recently, we are making torchaudio more lean, and we don't see a wide adoption of kaldi_pitch feature, so we decided to remove them.

See some of the discussion https://github.com/pytorch/audio/issues/1269

Pull Request resolved: https://github.com/pytorch/audio/pull/3368

Differential Revision: D46406176

Pulled By: mthrok

fbshipit-source-id: ee5e24d825188f379979ddccd680c7323b119b1e
      5bbbb1d5
    • moto's avatar
      Update data augmentation tutorial (#3375) · 2ba36b47
      moto authored 
      Summary:
Replace sox_effects with `torchaudio.io.AudioEffector`

1. To show case the new and better feature
2. To prepare for the upcoming removal of file-like support object

Pull Request resolved: https://github.com/pytorch/audio/pull/3375

Reviewed By: nateanl

Differential Revision: D46379016

Pulled By: mthrok

fbshipit-source-id: 70f24b62494204949f327f6ac6c49f315c9ee315
      2ba36b47
  59. 31 May, 2023 1 commit
  61. 26 May, 2023 2 commits
    • atalman's avatar
      Revert "Upgrade to FFmpeg5 (#3298)" (#3377) · 37779ef9
      atalman authored 
      Summary:
This reverts commit d38a7854.

This is temporary revert to unblock unit test migration from circleci to github

Pull Request resolved: https://github.com/pytorch/audio/pull/3377

Reviewed By: mthrok

Differential Revision: D46230498

Pulled By: atalman

fbshipit-source-id: 000d8a9ca00750fc1ca61f4c2cdd6e930a5ce46d
      37779ef9
    • Lakshmi Krishnan's avatar
      Improve RNN-T streaming decoding (#3295) · 9fc0dcaa
      Lakshmi Krishnan authored 
      Summary:
This commit fixes the following issues affecting streaming decoding quality
1. The `init_b` hypothesis is only regenerated from blank token if no initial hypotheses are provided.
2. Allows the decoder to receive top-K hypothesis to continue decoding from, instead of using just the top hypothesis at each decoding step.  This dramatically affects decoding quality especially for speech with long pauses and disfluencies.
3. Some minor errors regarding shape checking for length.

This also means that the resulting output is the entire transcript up until that time step, instead of just the incremental change in transcript.

Pull Request resolved: https://github.com/pytorch/audio/pull/3295

Reviewed By: nateanl

Differential Revision: D46216113

Pulled By: hwangjeff

fbshipit-source-id: 8f7efae28dcca4a052f434ca55a2795c9e5ec0b0
      9fc0dcaa