-
Notifications
You must be signed in to change notification settings - Fork 91
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
[BUG] dask scheduler not working: dask_cuda/cli.py AttributeError: 'function' object has no attribute 'command' #1179
Comments
This seems like an issue in |
This looks really strange. I'm also going to move to the dask-cuda issue tracker |
This is the first time I'm installing RAPIDS/cuDF with pip, but I created a new environment with Python 3.8 installing the packages as follows:
Then ran the scheduler successfully:
@pyotr777 could you try creating a new minimal environment like the above and see if that works? |
I also installed in a new env and was unable to reproduce |
In clean virtual environment installed as the following:
Install completed, but displayed an error:
I continued installation:
I confirm that dask scheduler starts without errors.
I tried to update numba.
The last installation finished with errors, however, now |
I don't know exactly why that error happened to you, but since I tried to create a new environment and got no install errors either. I know the cuDF PR to support NumPy>=1.24 was merged yesterday, but it doesn't pick NumPy>=1.24 for me either. To be honest I don't know exactly where the pip packages stand regarding that, hopefully this can be answered by @bdice or @vyasr , but at this point it doesn't seem like a Dask-related issue anymore, and perhaps should be transferred back to cuDF. |
To resolve this, you'll need to use a compatible numpy version, less than 1.24. The cuDF 23.06 release will support numpy 1.24. |
Ah right, that means NumPy>=1.24 was already installed when attempting to install cuDF, I didn't realize that. So a "clean environment" should have Python and PIP only, nothing else, not even NumPy to ensure cuDF will pick the correct supported versions of packages. |
pip's solver is limited and can't always fix issues such as with preinstalled packages. Manually downgrading the numpy version here should fix the issue, though. |
This is most likely automatically resolved in 23.06 and up, tentatively closing but feel free to reopen should this still be a problem. |
Describe the bug
After installing cudf cannot start dask scheduler.
Steps/Code to reproduce bug
Expected behavior
Dask scheduler works as usual.
Environment overview (please complete the following information)
Environment location: [Bare-metal, Docker, Cloud(specify cloud provider)]
Bare metal, Ubuntu 20.04.2 LTS, Python 3.8.10
Installed with
pip3 install cudf-cu11 --extra-index-url=https://pypi.nvidia.com
Environment details
Additional context
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered: