GitLab

* feat: efficient backprop for a_coeffs

* update lfilter docstring

* doc: double quotation

* use dafx paper for reference as it includes more details

* revise docstring

Summary:
Pull Request resolved: https://github.com/pytorch/audio/pull/3803

The model checkpoint path can not be created for Squim models. Use the latest download_asset method to fix it.

Reviewed By: moto-meta

Differential Revision: D59061348

Differential Revision: D57294298

Pull Request resolved: https://github.com/pytorch/audio/pull/3791

[codemod] c10::optional -> std::optional in pyspeech/experimental/csrc/decoders/TransducerDecoder.h +20

Differential Revision: D57294284

Pull Request resolved: https://github.com/pytorch/audio/pull/3793

[codemod] c10::optional -> std::optional in pytorch/audio/src/libtorio/ffmpeg/stream_reader/stream_processor.h +20

Differential Revision: D57294285

Pull Request resolved: https://github.com/pytorch/audio/pull/3792

Unpin mkl version and install pybind11 to get the windows CI working again

This fixes #3767

* Update tacotron2_pipeline_tutorial.py

- Fixed typo
- Clarified what was being done in different sections

Differential Revision: D54263224

Pull Request resolved: https://github.com/pytorch/audio/pull/3751

The lengths of targets and log_probs should be reversed.

Differential Revision: D53606067

Pull Request resolved: https://github.com/pytorch/audio/pull/3740

* add golf and dynonet paper

* doc: add references

* add EOF

* fix: line too long

* remove line end space

* remove indentation
Co-authored-by: moto <855818+mthrok@users.noreply.github.com>

---------
Co-authored-by: moto <855818+mthrok@users.noreply.github.com>

Summary:
This is not needed anymore after https://github.com/pytorch/test-infra/pull/4865.


Reviewed By: malfet, jeanschmidt, clee2000, NicolasHug

Differential Revision: D52735187

Pulled By: huydhn

* Update doc

* Update citation

Addresses #3688 

Back port from release/2.1 branch.
Need to git-fetch source code to get the version number dynamically

global audio backend is removed thus this is no-op.

PyTorch lightening is having issue with the nightly PyTorch.
Let the other tests still run.

When the input is zero Tensor, the result should be empty.

Differential Revision: D50696105

Pull Request resolved: https://github.com/pytorch/audio/pull/3682

Differential Revision: D50677606

Pull Request resolved: https://github.com/pytorch/audio/pull/3681

Differential Revision: D50633306

Pull Request resolved: https://github.com/pytorch/audio/pull/3675

Add torio top-level directory. It's not part of the package yet.