Update to Jupyterlab 4.0.10

[kallegrens]

GPU-Jupyter Feature Request

Thank you for your interest in enhancing GPU-Jupyter. We appreciate your ideas and feedback. Please fill out this brief form to help us understand your feature request.

Feature Description

Title
Update to Jupyterlab 4.1

Detailed Description

JupyterLab is now faster, thanks to improvements such as CSS rules optimization, CodeMirror 6, MathJax 3, and notebook windowing. By rendering only the parts of a notebook that fit in the web browser viewport, JupyterLab 4 is much more efficient than JupyterLab 3 was when working with large notebooks.
– Jupyter blog

Benefits
For the same reasons mentioned when deciding to upgrade to Jupyterlab 3.x in #44

Possible Implementation

Additional Information

Screenshots/Illustrations

References
v4.0 changelog
Jupyter blog

We value your contribution and look forward to exploring new features for GPU-Jupyter. Thank you for your support!

[kallegrens]

Nowhere in the generated Dockerfile is juptyterlab==3.6.6 mentioned. However, there are two extensions which don't yet support juptyterlab 4+. These are jupyterlab-drawio==0.9.0 and jupyterlab-spellchecker==0.7.3. However, the newer version 0.8.3 of spellchecker supports jupyterlab 4+.

I successfully built my own image with jupyterlab 4, simply without jupyterlab-drawio and with the updated spellchecker.

jupyterlab-drawio seems no longer to be maintained. The actively maintained ipydrawio does not yet support it either. The issue tracking this is: deathbeds/ipydrawio#119.

[ChristophSchranz]

Hi @kallegrens,
thanks for the suggestion and info! The GPU-Jupyter project inherits jupyterlab-related changes from the docker-stacks and will build images with Jupyterlab 4.X as soon as docker-stacks supports them and is stable enough for the update.

I've subscribed to deathbeds/ipydrawio#119 and will build new images if there is a solution or if there won't be any without this package.

[mkarikom]

Just wondering if I messed up my image build somehow?
I'm currently on v1.6_cuda-12.0_ubuntu-22.04 but it still seems to be built with jupyterlab==3.6.7
Regarding the comments above, I noticed that currently docker-stacks[main] seems to be making images with 4.1.3.

Admittedly, I'm pretty new to the 'stacks-based images, not entirely clear on how these images work vs those pulled by vanilla docker-compose, so I probably messed up something basic.

[ChristophSchranz]

Hi @mkarikom
The image v1.6_cuda-12.0_ubuntu-22.04 was built based on the commit 1494233e27cdc70e3766ea2518e7153ee425fc4f of docker-stacks. This one comes with jupyterlab==3.6.7.
It is good to hear that docker-stacks now supports Jupyterlab version 4.X.

I will check the Jupyterlab version and also the PyTorch compatibility with CUDA 12.3 (issue #140) and then build a new image with an updated docker-stacks hash.

[ChristophSchranz]

Unfortunately, drawio still doesn't support JupyterLab 4.X.

(base) jovyan@d07e814f66eb:~/work/Getting_Started$ conda install -c conda-forge jupyterlab-drawio
Channels:
 - conda-forge
 - nvidia
Platform: linux-64
Collecting package metadata (repodata.json): done
Solving environment: done

## Package Plan ##

  environment location: /opt/conda

  added / updated specs:
    - jupyterlab-drawio


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    aiofiles-22.1.0            |     pyhd8ed1ab_0          17 KB  conda-forge
    aiosqlite-0.19.0           |     pyhd8ed1ab_0          19 KB  conda-forge
    jupyter_client-7.4.9       |     pyhd8ed1ab_0          97 KB  conda-forge
    jupyter_server_fileid-0.9.1|     pyhd8ed1ab_0          20 KB  conda-forge
    jupyter_server_ydoc-0.8.0  |     pyhd8ed1ab_0          15 KB  conda-forge
    jupyter_ydoc-0.2.4         |     pyhd8ed1ab_0          78 KB  conda-forge
    jupyterlab-3.6.7           |     pyhd8ed1ab_0         5.4 MB  conda-forge
    jupyterlab-drawio-0.9.0    |     pyhd8ed1ab_0         624 KB  conda-forge
    jupyterlab-git-0.44.0      |     pyhd8ed1ab_0         531 KB  conda-forge
    jupyterlab_pygments-0.3.0  |     pyhd8ed1ab_0          18 KB  conda-forge
    nbdime-3.2.1               |     pyhd8ed1ab_0         3.8 MB  conda-forge
    notebook-6.5.6             |     pyha770c72_0         299 KB  conda-forge
    pyzmq-24.0.1               |  py311ha4b6469_1         573 KB  conda-forge
    y-py-0.5.9                 |  py311hfe55011_0         1.2 MB  conda-forge
    ypy-websocket-0.8.2        |     pyhd8ed1ab_0          17 KB  conda-forge
    ------------------------------------------------------------
                                           Total:        12.7 MB

The following NEW packages will be INSTALLED:

  aiofiles           conda-forge/noarch::aiofiles-22.1.0-pyhd8ed1ab_0 
  aiosqlite          conda-forge/noarch::aiosqlite-0.19.0-pyhd8ed1ab_0 
  jupyter_server_fi~ conda-forge/noarch::jupyter_server_fileid-0.9.1-pyhd8ed1ab_0 
  jupyter_server_yd~ conda-forge/noarch::jupyter_server_ydoc-0.8.0-pyhd8ed1ab_0 
  jupyter_ydoc       conda-forge/noarch::jupyter_ydoc-0.2.4-pyhd8ed1ab_0 
  jupyterlab-drawio  conda-forge/noarch::jupyterlab-drawio-0.9.0-pyhd8ed1ab_0 
  y-py               conda-forge/linux-64::y-py-0.5.9-py311hfe55011_0 
  ypy-websocket      conda-forge/noarch::ypy-websocket-0.8.2-pyhd8ed1ab_0 

The following packages will be DOWNGRADED:

  jupyter_client                         8.6.1-pyhd8ed1ab_0 --> 7.4.9-pyhd8ed1ab_0 
  jupyterlab                             4.1.5-pyhd8ed1ab_0 --> 3.6.7-pyhd8ed1ab_0 
  jupyterlab-git                        0.50.0-pyhd8ed1ab_1 --> 0.44.0-pyhd8ed1ab_0 
  jupyterlab_pygmen~                     0.3.0-pyhd8ed1ab_1 --> 0.3.0-pyhd8ed1ab_0 
  nbdime                                 4.0.1-pyhd8ed1ab_0 --> 3.2.1-pyhd8ed1ab_0 
  notebook                               7.1.2-pyhd8ed1ab_0 --> 6.5.6-pyha770c72_0 
  pyzmq                              25.1.2-py311h34ded2d_0 --> 24.0.1-py311ha4b6469_1 


Proceed ([y]/n)?

I suggest closing this issue with the support of Juypterlab 4.1, as drawio seems to be not actively maintained at the moment.

[ChristophSchranz]

Closing with #144