2i2c-org / utoronto-image

User image for the UToronto Hub

Image for jupyter.utoronto.ca

No more new R packages in this image

We are moving users of R to the R Hub, which operates off the UToronto R image. R will be eventually removed from here. No more new R packages will be added here.

utoronto-image's Issues

Move to using repo2docker files only

This was a groundbreaking image when we first set it up, primarily for its R functionality. However, enough have been merged into repo2docker now that I think we can migrate to using that for ease of maintenance.

We should coordinate with utoronto folks to make sure this isn't disruptive. Might mean we align this on their school holiday boundaries in a few months.

• Move to python 3.10 in environment.yml file
• Move to R 4.2 (latest version as I speak)
• Move all packages in install.R to their latest versions
• Bump up versions of all python dependencies in environment.yml to their latest versions
• Move all java, linux desktop installations to apt.txt
• Move bioconductor install back into install.R
• Move R CMD javareconf and the textblob.download_corpora to a postBuild file

Request to install Tensorflow and Keras for Python and R

This is a request coming from a Freshdesk support ticket: https://2i2c.freshdesk.com/a/tickets/65
We had a conversation in Slack about this one and we decided we are going to implement the update (instead of the user itself) because of the inherent complexity at the time to install those libraries.
The timeline and specific versions are not clear yet but I already have requested that information.

r2d-action broken?

Referencing PR #19

• The image build succeeded in the PR
• But it fails on push to main

I haven't tracked down the actual error message yet since the logs from the r2d-action are very long and verbose.

Is there a known issue regarding not being able to push this image to quay.io from CI/CD?

cc: @yuvipanda @damianavila

Request to install new python libraries

Context

This is coming from a support ticket: https://2i2c.freshdesk.com/a/tickets/114.

We previously had an issue with TF-related customizations, but when those involved RStudio: https://discourse.jupyter.org/t/openssl-mismatch-between-rstudio-and-conda-environments/14123

In this case, the requested confirmed:

For me this is just for the python environment. No RStudio requests for me!

So, hopefully, it should not be that difficult.

We have discussed this issue in the following Slack thread: https://2i2c.slack.com/archives/C028WU9PFBN/p1657296974240369

Proposal

Install the following python packages:

• pymc3
• tensorflow_probability

Updates and actions

In the Slack thread we agreed on:

1. create an issue with the request in the utoronto image repo [DONE]
2. try to update the image in a timebox manner (no more than 1-2 days)
3. we update the ticket letting them know it is done (or if we failed)
4. we remember them any image update should be championed by the comm representative in the short term future and with the expectation they should manage their image in the medium term future.

Reduce image build time

51 min to build the image... wondering if we can reduce that somehow 🤔

Originally posted by @damianavila in #3 (review)

Request to install python package textblob

Context

Origin: https://2i2c.freshdesk.com/a/tickets/187

Can you please install the latest version of Python package textblob on the U of T JupyterHub?

Proposal

Add textblob to environment.yml. This package is available in conda-forge and appear to be fairly stable.

According to the docs, there is some necessary "NLTK corpora" that needs to be downloaded as well. Presumably that is a "run once" kind of thing that could be done when the image is being built.

Updates and actions

• Add package to environment.yml
• Download the NLTK corpora as part of building the image
• Verify the package works as expected (example workflow )
• Close ticket on freshdesk

Request to install new python libraries

Context

Support Request via: https://2i2c.freshdesk.com/a/tickets/497

Proposal

Install the below (with approriate pinning)

contextily
ibpysal
esda
splot
xlrd
numba==0.53

Updates and actions

No response

Activate conda environment before starting notebook server

From https://2i2c.freshdesk.com/a/tickets/594, they are getting errors about PROJ environment variables not being set. This is because the conda environment is not being activated before the notebook server starts. The PATH env vars we set account for most of what conda activate does, but turns out conda activate also sets environment variables!

So we should properly activate the conda environment before starting the notebook server.

Migrate to JupyterLab3

We should upgrade the user image to use JupyterLab3.

• Original issue ➡️ utoronto-2i2c/jupyterhub-deploy#81
• Existing PR addressing this in the old utoronto deployment repository ➡️ utoronto-2i2c/jupyterhub-deploy#98
• Freshdesk ticket request ➡️ https://2i2c.freshdesk.com/a/tickets/57

Add PyMC and TensorFlow Probability to the image

Context

• Requested in https://2i2c.freshdesk.com/a/tickets/1259

Not sure if these kinds of large packages require a separate, dedicated ML image?

Proposal

No response

Updates and actions

No response