Skip to content

Add Windows CI#394

Merged
vincentqb merged 8 commits into
pytorch:masterfrom
peterjc123:win_fix
Jan 8, 2020
Merged

Add Windows CI#394
vincentqb merged 8 commits into
pytorch:masterfrom
peterjc123:win_fix

Conversation

@peterjc123

@peterjc123 peterjc123 commented Jan 6, 2020

Copy link
Copy Markdown
Contributor

No description provided.

@peterjc123 peterjc123 changed the title [WIP] Add Windows CI Add Windows CI Jan 6, 2020
@peterjc123

peterjc123 commented Jan 6, 2020

Copy link
Copy Markdown
Contributor Author

cc @vincentqb Should be ready. However, no tests are running at current.

@vincentqb vincentqb left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for working on this!

However, no tests are running at current.

Have you tried running them at all? Of course, a few depend on sox and will fail.

Comment thread torchaudio/_backend.py


_audio_backend = "sox"
_audio_backend = "soundfile" if platform.system() == "Windows" else "sox"

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You recommend making soundfile the default in windows? If that is the case, I'm thinking we may want to make soundfile the default everywhere. Since we are doing the 0.4.0 release now, it may be a good idea to bring this in now so that we have window supported.

Thoughts?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You recommend making soundfile the default in windows?

I made the change because I was not able to compile the sox backend on Windows. So actually, soundfile is the only available choice for Windows.

If that is the case, I'm thinking we may want to make soundfile the default everywhere.

The major limitation of soundfile is that it cannot read the file with the MP3 format, which is possible in sox.

@peterjc123 peterjc123 Jan 7, 2020

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Another difficulty I met is that there are no available pysoundfile packages hosted on Anaconda Cloud for Windows. Although the current one is marked as noarch, it is depending on a gcc package which is not available on Windows. So for conda users, the installation script will look a bit weird like this.

conda install -c pytorch torchaudio
pip install soundfile

@peterjc123

peterjc123 commented Jan 7, 2020

Copy link
Copy Markdown
Contributor Author

Have you tried running them at all?

Test results: https://gist.github.com/peterjc123/4f5b43ce372905c82333f7693a65a15c

@vincentqb vincentqb left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for doing this! Since sox is not an option on windows, we'll fall back to soundfile as you suggest.

@vincentqb vincentqb merged commit be5b2d5 into pytorch:master Jan 8, 2020
@peterjc123 peterjc123 deleted the win_fix branch January 9, 2020 05:50
@vincentqb vincentqb mentioned this pull request Feb 4, 2020
11 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants