Switch to Python 3.13

[mathbunnyru]

Describe your changes

Issue ticket if applicable

Checklist (especially for first-time contributors)

• [ ] I have performed a self-review of my code
• [ ] If it is a core feature, I have added thorough tests
• [ ] I will try not to use force-push to make the review process easier for reviewers
• [ ] I have updated the documentation for significant changes

[mathbunnyru]

I don't plan to merge this now, but let's see how many things are broken

[mathbunnyru]

We depend on updating mamba to v2: https://github.com/jupyter/docker-stacks/pull/2147

#15 9.104 error    libmamba Could not solve for environment specs
#15 9.104     The following packages are incompatible
#15 9.104     ├─ mamba <2.0.0 is installable with the potential options

[mathbunnyru]

If you are interested in making Python 3.13 work for these images, please help conda-forge to build missing packages, here is the current progress: https://conda-forge.org/status/migration/?name=python313

[mathbunnyru]

Python 3.13.1 has been released, so we're waiting for the packages to be compatible with 3.13. When tests pass, I will merge this, but I have no idea if it will be soon.

[mathbunnyru]

The upstream PR to build conda for python3.13 in conda-forge has been merged 🎉 https://github.com/conda-forge/conda-feedstock/pull/258

Let's see if everything else builds now (or if something fails)

[mathbunnyru]

Ok, there are 3 upstream issues:

1. TensorFlow doesn't support Python 3.13: https://github.com/tensorflow/tensorflow/issues/78774 It seems that no one answers when the support will be added, unfortunately.
2. rpy2: Waiting for https://github.com/conda-forge/rpy2-feedstock/pull/127, but more for https://github.com/conda-forge/rpy2-feedstock/pull/128
3. ~We install pinned version of pandas for spark-related images. v4.0.0-rc1 tag and GitHub release has been created, but no binaries are available on https://dlcdn.apache.org/spark/. This version already uses last pandas 2.2.3, which already works with python 3.13 in conda-forge. So, we're waiting for new binaries to become available.~

[mathbunnyru]

~spark is almost released, shouldn't be long: https://github.com/jupyter/docker-stacks/pull/2300~

[flaviomartins]

2. rpy2: Waiting for Rebuild for python 3.13 conda-forge/rpy2-feedstock#127, but more for rpy2 v3.5.17 conda-forge/rpy2-feedstock#128

Rebuild for python 3.13 merged. See: https://github.com/conda-forge/rpy2-feedstock/pull/132

[mathbunnyru]

2. rpy2: Waiting for Rebuild for python 3.13 conda-forge/rpy2-feedstock#127, but more for rpy2 v3.5.17 conda-forge/rpy2-feedstock#128

Rebuild for python 3.13 merged. See: https://github.com/conda-forge/rpy2-feedstock/pull/132

Thanks @flaviomartins! Awesome work 🔥 I actually wanted to try it out on the weekend, but let me restart it now.

[flaviomartins]

Thanks @flaviomartins! Awesome work 🔥 I actually wanted to try it out on the weekend, but let me restart it now.

The datascience image failed due to rpy2 requirements resolution. I uploaded a new build which seems to fix this https://github.com/conda-forge/rpy2-feedstock/pull/137.

[mathbunnyru]

Tensorflow is not yet capable with Python 3.13, but pip ens up downloading another version: Downloading tensorflow_cpu-2.20.0rc0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (4.5 kB)

https://github.com/tensorflow/tensorflow/releases/tag/v2.20.0-rc0

I hope it will be soon released and we will easily merge this

[flaviomartins]

Tensorflow is not yet capable with Python 3.13, but pip ens up downloading another version: Downloading tensorflow_cpu-2.20.0rc0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (4.5 kB)

https://github.com/tensorflow/tensorflow/releases/tag/v2.20.0-rc0

I hope it will be soon released and we will easily merge this

I was also surprised by the tensorflow images build success. While we are talking about this yesterday they released an RC with wheels for Python 3.13 👍

[anrayliu]

Ok, there are 3 upstream issues:

1. TensorFlow doesn't support Python 3.13: It doesn't support on python3.13 tensorflow/tensorflow#78774 It seems that no one answers when the support will be added, unfortunately.
2. rpy2: Waiting for Rebuild for python 3.13 conda-forge/rpy2-feedstock#127, but more for rpy2 v3.5.17 conda-forge/rpy2-feedstock#128
3. ~We install pinned version of pandas for spark-related images. v4.0.0-rc1 tag and GitHub release has been created, but no binaries are available on https://dlcdn.apache.org/spark/. This version already uses last pandas 2.2.3, which already works with python 3.13 in conda-forge. So, we're waiting for new binaries to become available.~

Hello, I noticed that rpy2 ticket 127 and ticket 128 are both now closed. Are we just waiting for TensorFlow now?

[mathbunnyru]

Ok, there are 3 upstream issues:

1. TensorFlow doesn't support Python 3.13: It doesn't support on python3.13 tensorflow/tensorflow#78774 It seems that no one answers when the support will be added, unfortunately.
2. rpy2: Waiting for Rebuild for python 3.13 conda-forge/rpy2-feedstock#127, but more for rpy2 v3.5.17 conda-forge/rpy2-feedstock#128
3. ~We install pinned version of pandas for spark-related images. v4.0.0-rc1 tag and GitHub release has been created, but no binaries are available on https://dlcdn.apache.org/spark/. This version already uses last pandas 2.2.3, which already works with python 3.13 in conda-forge. So, we're waiting for new binaries to become available.~

Hello, I noticed that rpy2 ticket 127 and ticket 128 are both now closed. Are we just waiting for TensorFlow now?

Yes, please see the comments above

[anrayliu]

Thanks for the quick reply. As a side note, how often do these Python upgrades happen (one month after each new version, only with major versions, every 2 months, etc)?

[mathbunnyru]

Thanks for the quick reply. As a side note, how often do these Python upgrades happen (one month after each new version, only with major versions, every 2 months, etc)?

We update major python version (like this PR does from 3.12 to 3.13) per our policy: https://jupyter-docker-stacks.readthedocs.io/en/latest/maintaining/tasks.html#updating-python-version

Images are rebuilt weekly automatically, so when a new patch release is available on conda-forge, our images will start using it automatically

[flaviomartins]

TF 2.20 has been released see: https://github.com/tensorflow/tensorflow/releases/tag/v2.20.0

[mathbunnyru]

TF 2.20 has been released see: tensorflow/tensorflow@v2.20.0 (release)

Awesome, thanks for letting me know, I updated the changelog date and the list of old images. When CI is green I will check release tensorflow is getting installed and will merge this.