GitLab

Co-authored-by: Vincent Quenneville-Belair <vincentqb@gmail.com>

All macOS CI jobs were failing at `brew`.
The issue was the macOS version `10.13`, which is no longer supported by `brew`.
The macOS version is indirectly controlled by `xcode` version as listed [here](https://circleci.com/docs/2.0/testing-ios/#supported-xcode-versions).

Updating the `xcode` to `"12.0"` as in PyTorch core bump macOS version to `10.15` and it resolves the issue.
`Python 3.9` is still failing but it's packaging issue on upstream, which needs to be resolved separately.

It was reported in #1478 that spectrogram masking operations were done in-place and modified the original input tensors. This PR fixes this behavior and adds tests to ensure that the input tensor is not changed.

Replace the prototype RNNT implementation (using warp-transducer) with one without external library dependencies

Should be moved back to stable once
https://github.com/pytorch/pytorch/issues/56654 is closed

- Put functional test logic into one place, `functional_impl.py`
- Tidy imports

`wavpack` is a format not supported/tested in torchaudio.
Leaving the option blank can cause the issue like #1461 in untested environment.

* Use torchaudio melscale instead of librosa

* Fixed floor_divide deprecation warnings seen in pytest output

* Fixed warning in test_flanger_triangle_linear

The VAD function trims the input tensor to the first instance of voice
activity on any channel or item. Trimming batches this way may be
undesirable as the item with earliest activity will dominate. Either
way, the batch behaviour does not match the itemwise behaviour.

The VAD batch consistency tests currently pass out of coincidence, but
they specify incorrect behaviour. This commit removes them.

Use shorter input sequences to avoid time out error on CI