Which version of python and cuda did you use?

Question

Which version of python and cuda did you use?

TheRealMephisto opened this issue 6 years ago · comments

Hi,
there seems to be a problem with using newer versions of python (3.7) and cuda (9.2).
I get two errors:
One complaining about nn.Container beeing deprecated, telling me to subclass nn.Module instead.
And another one complaining about using a different target size torch.size([128]) than the input size torch.size([128],[1]).
I think using the right versions of python and cuda should solve the problem.
So please tell me which you'd used, I didn't find it anywhere on the project Page here on Github.
Greetings!

Lukas Mosser · Answer 1 · Thu Nov 15 2018 05:51:03 GMT+0800 (China Standard Time)

Hi @TheRealMephisto!

I believe I used python 2.7 and cuda 8.0 for this.
The main issue here may be that I used an older version of Pytorch.
It was very likely Pytorch 0.2. I do not have a version that runs on version 1.0 but I believe there shouldn't be too much effort in porting. Main parts being changing nn.Containers to nn.Module.
I hope this helps, let me know if you run into more issues!

WechselWissen · Answer 2 · Mon Dec 03 2018 01:11:47 GMT+0800 (China Standard Time)

Thank you for your answer!
It worked for me with that versions, so I will leave it as it is for now.

WechselWissen · Answer 3 · Mon Dec 31 2018 06:02:48 GMT+0800 (China Standard Time)

Hi @LukasMosser !

Can you give me some more precise advice on how to port the network to newer versions?
I'm not sure what exactly to do...

Thanks in Advance!

Lukas Mosser · Answer 4 · Tue May 28 2019 00:41:35 GMT+0800 (China Standard Time)

Hi @TheRealMephisto
Are you still working on this?

WechselWissen · Answer 5 · Sat Jun 01 2019 20:53:22 GMT+0800 (China Standard Time)

Hi @LukasMosser ,
thank you for asking.
I'm currently not working on this.