Skip to content

GitLab

Switch branch/tag
  1. 23 Feb, 2023 2 commits
  3. 22 Feb, 2023 3 commits
  5. 21 Feb, 2023 1 commit
    • Chin-Yun Yu's avatar
      Fix contiguous error when backpropagating through lfilter (#3080) · 6ab1325a
      Chin-Yun Yu authored 
      Summary:
I encountered the following errors when using the filter with gradients being enabled.

```sh
Traceback (most recent call last):
  File "/home/ycy/working/audio/test_backward.py", line 20, in <module>
    loss.backward()
  File "/home/ycy/miniconda3/envs/nightly/lib/python3.10/site-packages/torch/_tensor.py", line 488, in backward
    torch.autograd.backward(
  File "/home/ycy/miniconda3/envs/nightly/lib/python3.10/site-packages/torch/autograd/__init__.py", line 197, in backward
    Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
RuntimeError: Expected input_signal_windows.is_contiguous() && a_coeff_flipped.is_contiguous() && padded_output_waveform.is_contiguous() to be true, but got false.  (Could this error message be improved?  If so, please report an enhancement request to PyTorch.)
```
This can happen if the outputs from lfilter was used by other operations.

### How to reproduce
The following script can reproduce the error on the stable and nightly versions.

```python
import torch
import torch.nn.functional as F
from torchaudio.functional import lfilter

a = torch.rand(250, 26, requires_grad=True)
b = torch.ones(250, 26, requires_grad=True)
x = torch.rand(250, 1024, requires_grad=True)
w = torch.eye(1024).unsqueeze(1)

y = lfilter(x, a, b, False)
y = F.conv_transpose1d(
    y.t().unsqueeze(0),
    w,
    stride=256,
).squeeze()
print(y.shape)
target = torch.ones_like(y)
loss = torch.nn.functional.mse_loss(y, target)
loss.backward()
```

### Cause

The inner call of differentiable IIR in the backward pass needs to ensure the input is contiguous. Adding a `contiguous()` call solve the problem.

Pull Request resolved: https://github.com/pytorch/audio/pull/3080

Reviewed By: xiaohui-zhang

Differential Revision: D43466612

Pulled By: mthrok

fbshipit-source-id: 375e0a147988656da47ac8397f7de6eae512a655
      6ab1325a
  7. 17 Feb, 2023 3 commits
    • hwangjeff's avatar
      Make lengths optional for speed functions and modules (#3072) · 5af309d3
      hwangjeff authored 
      Summary:
Makes lengths input optional for `torchaudio.functional.speed`, `torchaudio.transforms.Speed`, and `torchaudio.transforms.SpeedPerturbation`.

Pull Request resolved: https://github.com/pytorch/audio/pull/3072

Reviewed By: nateanl, mthrok

Differential Revision: D43371406

Pulled By: hwangjeff

fbshipit-source-id: ecb38bcc2bfff5c5a396a37eff238b22238e795a
      5af309d3
    • atalman's avatar
      Add py3.11 to windows nightly conda (#3071) · e663095c
      atalman authored 
      Summary:
Same as: https://github.com/pytorch/vision/pull/7263

Pull Request resolved: https://github.com/pytorch/audio/pull/3071

Reviewed By: weiwangmeta

Differential Revision: D43377741

Pulled By: atalman

fbshipit-source-id: 0dbe0aaa10b9a4bad713563e98642b1a65e9ac07
      e663095c
    • Daniel Walker's avatar
      Add precodition check for contiguous emissions tensor (#3074) · 06b1cc9d
      Daniel Walker authored 
      Summary:
This PR adds a precondition check to the `CTCDecoder` that raises a helpful exception when called on a noncontiguous emissions tensor.

Currently, noncontiguous tensors can be passed into the CTCDecoder, which in turn passes the tensors to the backing Flashlight C++ library and results in undefined behavior, since Flashlight requires the tensors to be laid out in contiguous memory. The following code demonstrates the problem:

```
import torch
from torchaudio.models.decoder import ctc_decoder

tokens = ['a', '-', '|']
decoder = ctc_decoder(lexicon=None, tokens=tokens)

emissions = torch.rand(len(tokens), 2)  # N x T contiguous
emissions = emissions.t()  # T x N noncontiguous

batch = emissions.unsqueeze(0)
result = decoder(batch)  # undefined behavior!!!
```

I stumbled on the issue accidentally when I noticed the decoder wasn't giving the expected results on my input only to realize, finally, that the tensor I had passed in was noncontiguous. In my case, Flashlight was iterating over unrelated segments of memory where it had expected to find a contiguous tensor. A precondition check will hopefully save others from making the same mistake.

Pull Request resolved: https://github.com/pytorch/audio/pull/3074

Reviewed By: nateanl, xiaohui-zhang

Differential Revision: D43376011

Pulled By: mthrok

fbshipit-source-id: 7c95aa8016d8f9f2d65b5b816a859b28ea4629f5
      06b1cc9d
  9. 16 Feb, 2023 5 commits
  11. 15 Feb, 2023 5 commits
  13. 14 Feb, 2023 4 commits
  15. 11 Feb, 2023 1 commit
  17. 10 Feb, 2023 1 commit
  19. 09 Feb, 2023 3 commits
  21. 08 Feb, 2023 4 commits
    • moto's avatar
      Update the guard mechanism for FFmpeg-related features (#3028) · 98b3ac17
      moto authored 
      Summary:
Instead of raising an error when lazy import happens, this method allows to import features, and raises an error when the feature is being used.

This makes it easy to adopt the same error mechanism across different modules. It is how it's done for sox-related features.

Pull Request resolved: https://github.com/pytorch/audio/pull/3028

Reviewed By: xiaohui-zhang

Differential Revision: D42966976

Pulled By: mthrok

fbshipit-source-id: 423dfe0b8a3970cd07f20e841c794c7f2809f993
      98b3ac17
    • moto's avatar
      Build doc on GHA (#3043) · a0f8af4b
      moto authored 
      Summary:
The first step to migrate doc build to GHA.

Pull Request resolved: https://github.com/pytorch/audio/pull/3043

Reviewed By: xiaohui-zhang

Differential Revision: D43110816

Pulled By: mthrok

fbshipit-source-id: 91de5f3ac567188e7030f14c2827a202a1901f1a
      a0f8af4b
    • moto's avatar
      Suppres warning about archive timestamp (#3044) · b4c66d1f
      moto authored 
      Summary:
Currently, for each third party library checked out with ExternalProject_Add, the following warning is shown.

This commit set the policy so that the warning is not shown.

```
CMake Warning (dev) at ci_env/lib/python3.10/site-packages/cmake/data/share/cmake-3.25/Modules/ExternalProject.cmake:3075 (message):
  The DOWNLOAD_EXTRACT_TIMESTAMP option was not given and policy CMP0135 is
  not set.  The policy's OLD behavior will be used.  When using a URL
  download, the timestamps of extracted files should preferably be that of
  the time of extraction, otherwise code that depends on the extracted
  contents might not be rebuilt if the URL changes.  The OLD behavior
  preserves the timestamps from the archive instead, but this is usually not
  what you want.  Update your project to the NEW behavior or specify the
  DOWNLOAD_EXTRACT_TIMESTAMP option with a value of true to avoid this
  robustness issue.
```

Pull Request resolved: https://github.com/pytorch/audio/pull/3044

Reviewed By: xiaohui-zhang

Differential Revision: D43110818

Pulled By: mthrok

fbshipit-source-id: d2e20c9fdbbeeedb5ad546fe32dbda28c5bdd431
      b4c66d1f
    • DanilBaibak's avatar
      Switch to Nova MacOS Conda (#2908) · de54d864
      DanilBaibak authored 
      Summary:
Switch to Nova M1 Conda

Pull Request resolved: https://github.com/pytorch/audio/pull/2908

Reviewed By: seemethere, osalpekar

Differential Revision: D43093605

Pulled By: DanilBaibak

fbshipit-source-id: 9e44f26cfb87e277c3808ee59f50218b4629e86e
      de54d864
  23. 07 Feb, 2023 2 commits
  25. 06 Feb, 2023 1 commit
  27. 04 Feb, 2023 2 commits
  29. 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
  31. 02 Feb, 2023 2 commits