Skip to content

GitLab

Switch branch/tag
  1. 23 May, 2023 1 commit
  3. 22 May, 2023 1 commit
  5. 20 May, 2023 1 commit
  7. 04 May, 2023 1 commit
    • Xiaohui Zhang's avatar
      Extend mask_along_axis{,_iid} (#3289) · 74bd971a
      Xiaohui Zhang authored 
      Summary:
(1/2 of the previous [PR](https://github.com/pytorch/audio/pull/2360) which I accidentally closed)

The previous way of doing SpecAugment via Frequency/TimeMasking transforms has the following problems:
- Only zero masking can be done; masking by mean value is not supported.
- mask_along_axis is hard-coded to mask the 1st dimension and mask_along_axis_iid is hard-code to mask the 2nd or 3rd dimension of the input tensor.
- For 3D spectrogram tensors where the first dimension is batch or channel, features from the same batch or different channels have to use the same mask, because mask_along_axis_iid only support 4D tensors, because of the above hard-coding
- For 2D spectrogram tensors w/o a batch or channel dimension, Time/Frequency masking can't be applied at all, since mask_along_axis only support 3D tensors, because of the above hard-coding.
- It's not straightforward to apply multiple time/frequency masks by the current design.

To solve these issues, here we
- Extend mask_along_axis_iid to support 3D tensors and mask_along_axis to support 2D tensors. Now both of them are able to mask one of the last two dimensions (where the time or frequency dimension lives) of the input tensor.

The introduction of SpecAugment transform will be done in another PR.

Pull Request resolved: https://github.com/pytorch/audio/pull/3289

Reviewed By: hwangjeff

Differential Revision: D45460357

Pulled By: xiaohui-zhang

fbshipit-source-id: 91bf448294799f13789d96a13d4bae2451461ef3
      74bd971a
  9. 07 Mar, 2023 1 commit
  11. 25 Feb, 2023 1 commit
  13. 17 Feb, 2023 1 commit
  15. 15 Feb, 2023 1 commit
  17. 03 Feb, 2023 1 commit
    • moto's avatar
      Add Linux GPU unit tests on GHA (#3029) · 6bdd3830
      moto authored 
      Summary:
Add GitHub Action-based GPU test jobs.
- It seems that there is 2 hour upper cap so only running CUDA/GPU tests.
- Since Kaldi related features are not available, they are disabled.

Pull Request resolved: https://github.com/pytorch/audio/pull/3029

Reviewed By: hwangjeff

Differential Revision: D42983800

Pulled By: mthrok

fbshipit-source-id: 47fefe39c635d1c73ad6799ddacefd2666fe5403
      6bdd3830
  19. 26 Jan, 2023 1 commit
  21. 24 Jan, 2023 1 commit
  23. 06 Jan, 2023 1 commit
  25. 16 Dec, 2022 1 commit
    • Caroline Chen's avatar
      Rename resampling_method options (#2922) · e6bebe6a
      Caroline Chen authored 
      Summary:
resolves https://github.com/pytorch/audio/issues/2891

Rename `resampling_method` options to more accurately describe what is happening. Previously the methods were set to `sinc_interpolation` and `kaiser_window`, which can be confusing as both options actually use sinc interpolation methodology, but differ in the window function used. As a result, rename `sinc_interpolation` to `sinc_interp_hann` and `kaiser_window` to `sinc_interp_kaiser`. Using an old option will throw a warning, and those options will be deprecated in 2 released. The numerical behavior is unchanged.

Pull Request resolved: https://github.com/pytorch/audio/pull/2922

Reviewed By: mthrok

Differential Revision: D42083619

Pulled By: carolineechen

fbshipit-source-id: 9a9a7ea2d2daeadc02d53dddfd26afe249459e70
      e6bebe6a
  27. 14 Nov, 2022 1 commit
  29. 10 Nov, 2022 1 commit
  31. 08 Nov, 2022 1 commit
    • Caroline Chen's avatar
      Enable log probs input for rnnt loss (#2798) · ca478823
      Caroline Chen authored 
      Summary:
Add `fused_log_softmax` argument (default/current behavior = True) to rnnt loss.

If setting it to `False`, call `log_softmax` on the logits prior to passing it in to the rnnt loss function.

The following should produce the same output:
```
rnnt_loss(logits, targets, logit_lengths, target_lengths, fused_log_softmax=True)
```

```
log_probs = torch.nn.functional.log_softmax(logits, dim=-1)
rnnt_loss(log_probs, targets, logit_lengths, target_lengths, fused_log_softmax=False)
```

testing -- unit tests + get same results on the conformer rnnt recipe

Pull Request resolved: https://github.com/pytorch/audio/pull/2798

Reviewed By: xiaohui-zhang

Differential Revision: D41083523

Pulled By: carolineechen

fbshipit-source-id: e15442ceed1f461bbf06b724aa0561ff8827ad61
      ca478823
  33. 03 Aug, 2022 1 commit
    • bshall's avatar
      An implemenation of the ITU-R BS.1770-4 loudness recommendation (#2472) · 946b180a
      bshall authored 
      Summary:
I took a stab at implementing the ITU-R BS.1770-4 loudness recommendation (closes https://github.com/pytorch/audio/issues/1205). To give some more details:
- I've implemented K-weighting following csteinmetz1 instead of BrechtDeMan since it fit well with torchaudio's already implemented filters (`treble_biquad` and `highpass_biquad`).
- I've added four audio files to test compliance with the recommendation. These are linked in [this pdf](https://www.itu.int/dms_pub/itu-r/opb/rep/R-REP-BS.2217-2-2016-PDF-E.pdf). There are many more test files there but I didn't want to bog down the assets directory with too many files. Let me know if I should add or remove anything.
- I've kept many of the constant internal to the function (e.g. the block duration, overlap, and the absolute threshold gamma). I'm not sure if these should be exposed in the signature.
- I've implemented support for up to 5 channels (following both csteinmetz1 and BrechtDeMan). The recommendation includes weights for up to 24 channels. Is there any convention for how many channels to support?

I hope this is helpful! looking forward to hearing from you.

Pull Request resolved: https://github.com/pytorch/audio/pull/2472

Reviewed By: hwangjeff

Differential Revision: D38389155

Pulled By: carolineechen

fbshipit-source-id: fcc86d864c04ab2bedaa9acd941ebc4478ca6904
      946b180a
  35. 28 Jul, 2022 1 commit
  37. 23 Jun, 2022 1 commit
  39. 03 Jun, 2022 1 commit
  41. 01 Jun, 2022 1 commit
  43. 23 May, 2022 1 commit
    • Zhaoheng Ni's avatar
      Add assertion checks to multi-channel functions (#2401) · 38e530d7
      Zhaoheng Ni authored 
      Summary:
- The multi-channel functions only support complex-valued tensors for spectrogram and PSD matrices.
- The mask can be real-valued or complex-valued, hence there is no explicit assertion for mask.
- The shape of input Tensors need to be verified before the computation. For example, the shape of PSD matrix must be `(..., freq, channel, channel)`, the shape of the mask must be `(..., freq, time)`, etc.
- The autograd unittest of `apply_beamforming` has wrong dimensions for beamform_weights detected by the assertion check. FIx it in this PR.

Pull Request resolved: https://github.com/pytorch/audio/pull/2401

Reviewed By: carolineechen

Differential Revision: D36597689

Pulled By: nateanl

fbshipit-source-id: 6ad1adebe3726851cc1d865650bdf177a98985f6
      38e530d7
  45. 15 May, 2022 1 commit
    • John Reese's avatar
      [codemod][usort] apply import merging for fbcode (8 of 11) · d62875cc
      John Reese authored 
      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
      d62875cc
  47. 10 May, 2022 1 commit
  49. 08 Apr, 2022 1 commit
    • moto's avatar
      Add devices/properties badges (#2321) · 72ae755a
      moto authored 
      Summary:
Add badges of supported properties and devices to functionals and transforms.

This commit adds `.. devices::` and `.. properties::` directives to sphinx.

APIs with these directives will have badges (based off of shields.io) which link to the
page with description of these features.

Continuation of https://github.com/pytorch/audio/issues/2316
Excluded dtypes for further improvement, and actually added badges to most of functional/transforms.

Pull Request resolved: https://github.com/pytorch/audio/pull/2321

Reviewed By: hwangjeff

Differential Revision: D35489063

Pulled By: mthrok

fbshipit-source-id: f68a70ebb22df29d5e9bd171273bd19007a81762
      72ae755a
  51. 26 Feb, 2022 1 commit
    • Zhaoheng Ni's avatar
      Add apply_beamforming to torchaudio.functional (#2232) · 9c56ffb4
      Zhaoheng Ni authored 
      Summary:
This PR adds ``apply_beamforming`` method to ``torchaudio.functional``.
The method employs the beamforming weight to the multi-channel noisy spectrum to obtain the single-channel enhanced spectrum.
The input arguments are the complex-valued beamforming weight Tensor and the multi-channel noisy spectrum.

Pull Request resolved: https://github.com/pytorch/audio/pull/2232

Reviewed By: mthrok

Differential Revision: D34474561

Pulled By: nateanl

fbshipit-source-id: 2910251a8f111e65375dfb50495b6a415113f06d
      9c56ffb4
  53. 25 Feb, 2022 5 commits
  55. 17 Feb, 2022 1 commit
    • Zhaoheng Ni's avatar
      Refactor batch consistency test in functional (#2245) · 9cf59e75
      Zhaoheng Ni authored 
      Summary:
In batch_consistency tests, the `assert_batch_consistency` method only accepts single Tensor, which is not applicable to some methods. For example, `lfilter` and `filtfilt` requires three Tensors as the arguments, hence they don't follow `assert_batch_consistency` in the tests.
This PR refactors the test to accept a tuple of Tensors which have `batch` dimension. For the other arguments like `int` or `str`, they are given as `*args` after the tuple.

Pull Request resolved: https://github.com/pytorch/audio/pull/2245

Reviewed By: mthrok

Differential Revision: D34273035

Pulled By: nateanl

fbshipit-source-id: 0096b4f062fb4e983818e5374bed6efc7b15b056
      9cf59e75
  57. 16 Feb, 2022 2 commits
    • Zhaoheng Ni's avatar
      Refactor torchscript consistency test in functional (#2246) · 87d79889
      Zhaoheng Ni authored 
      Summary:
In torchscript_consistency tests, the `func` in each test method only accepts one `tensor` as the argument, for the other arguments of `F.xyz` method, they need to be defined inside the `func`. If there is no `Tensor` argument in `F.xzy`, the tests use a `dummy` tensor which is not used anywhere. In this PR, we refactor ``_assert_consistency`` and ``_assert_consistency_complex`` to accept a tuple of inputs instead of just one `tensor`.

Pull Request resolved: https://github.com/pytorch/audio/pull/2246

Reviewed By: carolineechen

Differential Revision: D34273057

Pulled By: nateanl

fbshipit-source-id: a3900edb3b2c58638e513e1490279d771ebc3d0b
      87d79889
    • Zhaoheng Ni's avatar
      Add complex dtype support in functional autograd test (#2244) · eeba91dc
      Zhaoheng Ni authored 
      Summary:
In autograd tests, to guarantee the precision, the dtype of Tensors are converted to `torch.float64` if they are real. However, the complex dtype is not considered. This PR adds `self.complex_dtype` support to the inputs.

Pull Request resolved: https://github.com/pytorch/audio/pull/2244

Reviewed By: mthrok

Differential Revision: D34272998

Pulled By: nateanl

fbshipit-source-id: e8698a74d7b8d99ee0fcb5f5cb5f2ffc8c80b9b5
      eeba91dc
  59. 09 Feb, 2022 1 commit
    • hwangjeff's avatar
      Fix librosa calls (#2208) · e5d567c9
      hwangjeff authored 
      Summary:
Yesterday's release of librosa 0.9.0 made args keyword-only and changed default padding from "reflect" to "zero" for some functions. This PR adjusts callsites in our tutorials and tests accordingly.

Pull Request resolved: https://github.com/pytorch/audio/pull/2208

Reviewed By: mthrok

Differential Revision: D34099793

Pulled By: hwangjeff

fbshipit-source-id: 4e2642cdda8aae6d0a928befaf1bbb3873d229bc
      e5d567c9
  61. 29 Dec, 2021 1 commit
  63. 23 Dec, 2021 1 commit
  65. 04 Nov, 2021 1 commit
  67. 03 Nov, 2021 2 commits