-
-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Unable to load folders for labeling in napari-dlc #2338
Comments
Hi @Andrianarivelo, could you please let us know which version of napari you're using? You can do this by running The full output would be useful to us (so we can see which package versions are installed and being used). |
Hi, thanks for the response, here's what I get after
|
The root issue seems to be with I've found this post describing a similar issue, and some comments with ways to install I believe the easiest that might work would be running As long as you get this |
Thanks a lot, I fixed it with But then I got : I fixed it by updating libstdc++ with The GUI works and I can load folders now |
I wanted to follow up on this because I spent most of the weekend trying to get deeplabcut installed, and most of that time trying to deal with this 'swrast' issue. In my case I was trying to (and now succeeded in) getting DLC installed in an Ubuntu 20.04 LXC container with NVidia GPU passthrough (running on an Ubuntu 22.04 host). The .yaml conda install failed due to some tensorflow issues I couldn't solve - I also couldn't get docker running (apparently this works better in privileged container which mine are not). I ended up copying the following installation recipe #2140 (comment) which got me a lot closer to a working state. The last hurdle though was the following error message (which may be related to the way NVIDIA GPU passthrough works on LXC).
At this point the GUI starts but much of it is broken. The key to getting around this specific issue is was to delete the $CONDA_PREFIX/lib versions of libffi.so.7 and libffi.7.so and have them instead loaded in a different clunky way. I am including my entire install 'recipe' so someone else can save the time of having to start this process from scratch. I've now tested this on two separate Ubuntu 20.04 containers (on the same host) and have been able to run DLC using my NVIDIA GPU. Note that this is probably a result of a specific combination of issues (again, maybe LXC GPU passthrough) and I would certainly recommend trying the standard (conda yaml) install before this one. install cuda 11.8 and cudnn 8.7 (see: https://www.enablegeek.com/blog/setting-up-cuda-and-cudnn-on-ubuntu-20-04/ and https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html)
CUDNN:
Probably restart your container/machine at this point for good measure. 'Manually' install DLC in conda environment:
Note this is part of the trick for getting around the 'swrast' issue, the last part of this fix is below I doubt the order of these operations matters, but this in fact the order I did them in:
Continuing with setting up this environment:
To resolve the 'drawtype' bug: For using Jupyter Notebook: Last part of fixing the swrast issue (credit to this github commenter spinalcordtoolbox/spinalcordtoolbox#3927 (comment)):
Note that conda updates could end up recreating these problematic '.so' files, so the issue would come back - if anyone knows of a better solution (that they have actually validated) please let me know - I tried very many of the other suggestion I came across on this subject, including those in the previous comments on this issue page, but it was this combination that got things unstuck. Hopefully this saves someone a weekend sometime down the road. |
Hi, I am also having the same issue as describe by Andrianarivelo here. I am using DLC 2.3.4 on Ubuntu 20.04. However, when i am running napari --info, I get the following: OpenGL:
Screens:
Plugins:
Please suggest the required solution to get around this issue. |
Is there an existing issue for this?
Bug description
Hello,
I am struggling with the new labeling GUI, I upgraded DLC to version 2.3.5, and when I load config.yaml, images or h5 data file by dropping them directly onto the canvas or via the File menu absolutely nothing happens. I extracted frames before, and I used DLC 2.2 for years but we find the new GUI really annoying. Same when I do deeplabcut.label_frames(config_path), nothing happens and I get the same log output.
Operating System
Ubuntu 20.04.4
DeepLabCut version
dlc version 2.3.5
DeepLabCut mode
single animal
Device type
gpu Nvidia 3090/
Steps To Reproduce
In DLC environment
napari
drop a folder from DeepLabCut's project labeled-data
Relevant log output
Anything else?
No response
Code of Conduct
The text was updated successfully, but these errors were encountered: