Skip to content
This repository has been archived by the owner on Sep 2, 2024. It is now read-only.

Question: which version of Python was used? #13

Open
ptc-lexvandersluijs opened this issue Jun 25, 2021 · 3 comments
Open

Question: which version of Python was used? #13

ptc-lexvandersluijs opened this issue Jun 25, 2021 · 3 comments

Comments

@ptc-lexvandersluijs
Copy link

Hi,
I would love to do some experiments with this fascinating work, but I'm having trouble getting the code to run.

At first I tried on a MacBook Pro M1, using Python 3.9, which I expected to be a challenge. It went better than expected, but ran into trouble with the open3d and python-opencv packages (no compatible package found).

Ultimately I switched to the Rosetta (i386) emulation on the same computer, using Python 3.8, but still no luck with Open3D. Then I read somewhere that using Python 3.7.7 should work, but this resulted in a humongous log of compiler errors and warnings. Since I'm pretty new to Python, can you say which version you used at the time of creation? And which operating system and version?

@griegler
Copy link
Contributor

Seems to be mostly about Open3D with M1, I suggest you raise an issue with Open3D.
I used/tested it with a MacBook with an Intel i7, macOS Catalina.

@ptc-lexvandersluijs
Copy link
Author

That's a valid point, thanks.

In the meantime -good news- I've gotten the code to work on Windows 10, using the Windows Linux Subsystem feature https://seanthegeek.net/234/graphical-linux-applications-bash-ubuntu-windows/. I wasn't able to get the CUDA driver to work according to this instruction https://docs.nvidia.com/cuda/wsl-user-guide/index.html#installing-nvidia-drivers, but did manage to convince pytorch to only use the CPU.

To others who may be interested: after a lot of thrashing about, there's a good chance that the only lines that really need to be changed are these two at the top of config.py, just change "cuda" to "cpu":

train_device = "cuda:0"
eval_device = "cuda:0"

So my next step is to try it with my own SfM datasets.

And before I forget: thanks for making this available, it is a really very impressive work!

This issue can be closed for what I'm concerned.

@Asherchi
Copy link

I installed python==3.7 torch 1.7.0 and corresponde version of torch-scatter and so on.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants