GitLab

- TorchScript-able `Wav2Vec2Model` class
- Factory functions for three configurations presented in the paper 
  - `wav2vec2_base`
  - `wav2vec2_large`
  - `wav2vec2_large_lv60k`

Replacing the config.guess with a newer version to support newer hardware, such as Nvidia Jetson.

Obtained from: https://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD

See: https://www.gnu.org/software/gettext/manual/html_node/config_002eguess.html

* Add CUDA binary builds

* Add "cuda_version": "cpu" to doc build jobs

* Add required cu_versions to pkg_helpers

* Regenerate config.yml

* Remove `class FunctionalComplex` header accidentally re-introduced in #1490 

* Update VAD docstring and check for input shape length

* Update docstring in forward for transform

* Address review feedback: merge tests, update wording

Precomputing and caching the resampling kernel in transforms provides speed improvements for resample, but no longer handles the automatic device and dtype recognition and construction based on input waveform. This is BC-breaking if users do not manually move the transforms object to the correct device and dtype, in which case calls to resample will fail if the input waveform is on gpu, or not of float32 dtype. Precomputing the kernel additionally results in very minor precision differences from previous implementation.

Since the update of Xcode, some of HTTP tests are failing on macOS, due to `reqests.get` fails to establish a connection to local server. This change increase the wait time to 2 seconds before test starts.

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