Kernel hangs in "TCIA_PROSTATEx_Prostate_MRI_Anatomy_Model.ipynb"
KumoLiu opened this issue · 11 comments
Describe the bug
Kernel hangs after itkwidgets visualization in jupyter notebook
https://github.com/Project-MONAI/tutorials/blob/main/model_zoo/TCIA_PROSTATEx_Prostate_MRI_Anatomy_Model.ipynb
To Reproduce
- docker pull projectmonai/monai:latest
- create a new container
- run into container
pip install 'itkwidgets[notebook]>=1.0a49'
from monai.transforms import LoadImage
root_dir = "/opt/monai/workspace/Data"
data_dir = os.path.join(root_dir, "Task09_Spleen")
train_images = sorted(glob.glob(os.path.join(data_dir, "imagesTr", "*.nii.gz")))
out = LoadImage(ensure_channel_first=True)(train_images[0])
view(image=out)
Expected behavior
Kernal will not hangs after using view.
Screenshots
Additional context
- Look similar to this one:
InsightSoftwareConsortium/itkwidgets#688 - After install
itkwidgets[notebook]>=1.0a49, look like there will be a requirement compatibility issues for pandas, not sure whether it's related the issue.
Collecting pandas>=2.0.0 (from dask-image->ngff-zarr[dask-image]>=0.4.3->itkwidgets>=1.0a49->itkwidgets[notebook]>=1.0a49)
Downloading pandas-2.2.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (19 kB)
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
cudf 24.2.0 requires pandas<1.6.0dev0,>=1.3, but you have pandas 2.2.2 which is incompatible.
dask-cuda 24.2.0 requires pandas<1.6.0.dev0,>=1.3, but you have pandas 2.2.2 which is incompatible.
dask-cudf 24.2.0 requires pandas<1.6.0dev0,>=1.3, but you have pandas 2.2.2 which is incompatible.
- After install the itkwidgets and exiting the Jupyter notebook, trying to launch the notebook again leads to a configuration error. It appears that the installation of itkwidgets is causing a disruption in the configuration settings.
cc @kirbyju @aylward and @thewtex, in case you can shed any light on this issue, it would be greatly appreciated.
Also created a ticket here InsightSoftwareConsortium/itkwidgets#743
Thanks.
Also tried this notebook with pytorch container: nvcr.io/nvidia/pytorch:24.03-py3.
See the same issue.
https://github.com/InsightSoftwareConsortium/itkwidgets/blob/v1.0a50/examples/integrations/MONAI/transform_visualization.ipynb
Thank you for the detailed report on the itkwidget repo. Matt and Brianna are looking into it!
Hello 👋 ,
With itkwidgets 1.0a50, itkwidgets supports JupyterLab 4 and Jupyter Notebook 7. The Jupyter project made these versions major breaking changes. So, there are a few approaches:
a) Update the JupyterLab and Jupyter Notebook version
b) Set the itkwidgets dependency to itkwidgets==1.0a49 until a) can be performed.
After install itkwidgets[notebook]>=1.0a49, look like there will be a requirement compatibility issues for pandas, not sure whether it's related the issue.
This seems to be unrelated in the transitive dependency arena, but a patch is here:
Testing and feedback to see if it addresses your use cause are appreciated. 🙏 Thanks!
Hi @thewtex, thanks for the quick response.
I update the notebook version but the same issue occurred.
Step to reproduce:
docker run -it --ipc=host --rm --net host --gpus all nvcr.io/nvidia/pytorch:24.03-py3
pip install -U notebook
Then run this notebook: https://github.com/InsightSoftwareConsortium/itkwidgets/blob/v1.0a50/examples/integrations/MONAI/transform_visualization.ipynb
You can see from the screen shot after view even import numpy as not be executed. Also can see the output from view.
BTW, where can I find the requirement for the notebook, I didn't find it in the requirement. Perhaps I want to try the version which can run in notebook 6.
Thanks!
@KumoLiu thanks for the reproducible example!
Addressed here:
InsightSoftwareConsortium/itkwidgets#745
After it is released, it will be better to install itkwidgets in the containers so all javascript is loaded at startup.
BTW, where can I find the requirement for the notebook, I didn't find it in the requirement. Perhaps I want to try the version which can run in notebook 6.
The dependencies are here:
We keep them as small and light and universal as possible.
Hi @KumoLiu , these improvements have now been released in itkwidgets 1.0a51.
Please try:
docker run -it --ipc=host --rm --net host --gpus all nvcr.io/nvidia/pytorch:24.03-py3
pip install -U notebook
pip install 'itkwidgets[notebook]==1.0a51'
# Start the notebook
Hi @thewtex, thanks for the quick fix.
I tried with 1.0a51, and the same issue occurred.
In addition, after this command pip install 'itkwidgets[notebook]==1.0a51' and jupyter notebook can not even launched correctly.
@KumoLiu thanks for testing.
I also observed that error in my testing when I tried installing widgetsnbextension (a different, unrelated package), made for notebook==6 when notebook==7 was installed.
I am wondering if there is something else in your installation?
Could you please try:
docker run -it -p 8888:8888 --ipc=host --rm --net host --gpus all thewtex/monai-1694or simply build and run the following Dockerfile:
FROM nvcr.io/nvidia/pytorch:24.03-py3
RUN pip install -U notebook
RUN pip install 'itkwidgets[notebook]==1.0a51'
RUN curl -L -o /transform_visualization.ipynb https://raw.github.com/InsightSoftwareConsortium/itkwidgets/main/examples/integrations/MONAI/transform_visualization.ipynb
CMD jupyter notebook /transform_visualization.ipynb
?
You should see:
(there is an issue with the last cell, but I have a fix in progress for that).
(there is an issue with the last cell, but I have a fix in progress for that).
This was a side-effect of fixing the pandas dependency issue, but it has been addressed in itkwidgets 1.0a52.