GitLab

Also replaces `assert_allclose` with `assertEqual`.
Co-authored-by: krishnakalyan3 <skalyan@cloudera.com>

Co-authored-by: krishnakalyan3 <skalyan@cloudera.com>

Co-authored-by: krishnakalyan3 <skalyan@cloudera.com>

* improve commonvoice test

* update yesno test

* refactor commonvoice dataset unittest

Resolves #1125 where dataset metadata does not contain an extension.

* Import torchaudio #1105 37692d8d

Summary: Import torchaudio up to #1105 37692d8d

Reviewed By: datumbox

Differential Revision: D25671497

fbshipit-source-id: 5af11c801321f2bb964245ac6ed74979310f4b5f

The use of `walk_files` made it ambiguous who is responsible to locate 
the correct set of files. (Dataset class? or utility?)
In fact, just glob-ing everything is not the right problem being solved in implementing
Dataset, because if you have a specific dataset you consider to access, then
the directory structure and file locations are determined. No need to do arbitral number of recursions.
Each Dataset implementation should be glob-ing the right set of files it requires.

There's an issue with the CUDA 11.0 package for conda with
conda-package-handling that relates to https://github.com/conda/conda-package-handling/issues/71



This should fix issues with the conda smoke testing we encountered previously.
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

Co-authored-by: krishnakalyan3 <skalyan@cloudera.com>

Co-authored-by: krishnakalyan3 <skalyan@cloudera.com>

* Removes code for download logics 
* [BC-breaking] Changes the meaning of `root` argument to the exact directory of the dataset
* Deprecates the constructor arguments for download and subdirectory construction

* Resolves https://github.com/pytorch/audio/issues/1095

- yesno
- librispeech
- libritts
- speechcommands
Co-authored-by: krishnakalyan3 <skalyan@cloudera.com>
Co-authored-by: Vincent Quenneville-Belair <vincentqb@gmail.com>