Re: PyTorch failure after onnx update
FYI, if you somehow need a functional pytorch build:
$ dget http://debomatic-amd64.debian.net/debomatic/experimental/pool/pytorch_1.12.0~rc1-1/pytorch_1.12.0~rc1-1_amd64.changes
$ sudo debi -u pytorch_1.12.0~rc1-1_amd64.changes
It pulls some packages from experimental like opencv 4.6.0. And there is no longer
symbol not found error. Note, debomatic will automatically cleanup the builds...
So when it get expired, you may need to download it from the NEW queue if possible.
As for the experimental->unstable ... I'm afraid that will not happen soon.
I still have to deal with flatbuffers transition before that.
On Tue, 2022-06-28 at 09:53 +0200, Gard Spreemann wrote:
> "M. Zhou" <lumin@debian.org> writes:
>
> > No worries. I'm doing the PyTorch 1.12.X migration.
> > http://debomatic-amd64.debian.net/distribution#unstable/pytorch/1.12.0~rc1-1
> > And I'm aware of the break before uploading onnx to unstable.
> > I'm finalizing 1.12.0~rc1 and it will enter the NEW queue in several days.
> >
> > Due to some complicated reasons, I permanently lost access to any strong
> > building machine. All I have now is my I7-8565U CPU. In that sense the only
> > thing I can do is to do the reckless transition, uploading dependencies to
> > unstable directly so that I can abuse debomatic a little bit to finish
> > the transition.
> >
> > I have to upload them to unstable because there is no chance to use an
> > --extra-packages=... argument like sbuild.
>
> Understood. As always, thanks for all your hard work making Debian a
> solid platform also for deep learning!
>
> -- Gard
>
Reply to: