-
Notifications
You must be signed in to change notification settings - Fork 220
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
anaconda/python 3.11 on all supported platforms #13082
Comments
Hi @AntonOfTheWoods , |
This is puzzling. Python 3.11.1 has been released but we don't have any Python 3.11 in anaconda channel yet. I'm curious what makes Python 3.11 so difficult to support? Is it because there are much more 3rd party libraries you will need to build with Python 3.11? If so, can we build those libraries gradually? |
@alexlang74 , I'm just a user. There is nothing difficult about implementing this. In fact you can just add condaforge and use that instead. The maintainers don't need this and obviously don't care that we do, or at least don't have the free time to care! It's that simple. |
This seems to be a blocker for PyTorch vision library Python 3.11 support. |
dependencies. Conda depends about a lot libraries and conda-forge can be not stable. |
Sure, I am just helping to provide support for recent python versions to a FOSS project that uses anaconda. I would never use it myself, for this very reason. I will suggest that we remove anaconda completely but that takes time and effort... |
@AntonOfTheWoods I think this issue can be kept open. I understand the sentiment behind the statement "The maintainers don't need this and obviously don't care that we do, or at least don't have the free time to care! It's that simple" but in the reality is more nuanced than that. While PyPI packages have to maintain their own individual package compatibilities with a new Python release, Anaconda really is a complete ecosystem of packages which cross language boundaries. In many ways is better thought of in the way that Linux distributions are: an intermediate layer between the packages themselves and the system architecture. This implies trade-offs in the speed to build out distributions, and just shipping CPython by itself solves few problems because the ecosystem is built around targeting high-value toolchains like deep learning workflows which encapsulate a significant set of package requirements. I don't work on Anaconda directly but do maintain a set of package for conda and can attest to the level of effort this can entail. Conda-forge is great as a rolling release, automate everything environment but there are trade-offs in ABI compatibility and other areas that Anaconda makes. |
@scdub , so you think they do have the time and still don't even address the issue (here, for example)? |
Please re-open, it is not solved and important to many people (See the link above to pytorch/vision). This is currently also a blocker for the next numba release, see and |
If we are completely honest, I think this is a perfect example of why anaconda is a horrible solution for what many people are using it for. If you need a little reactivity in updates of python versions, then you need something else. I'll reopen and unsubscribe, as I don't need this anymore... |
Thank you for your patience. At Anaconda, we provide a cohesive distribution across the 7 architectures and 4 python versions we support, which does incur some delay in releasing. We are currently working on the build out of the distribution for python 3.11. |
When do you expect a release? At least to be able to create environments. |
Hi Johnny, Python 3.11 along with setuptools and pip have been pushed to defaults today. |
@cbouss some dependencies such as numpy and scipy still need to be updated. |
scipy is already updated! |
@ZupoLlask: Is it? Obviously both numpy and scipy both supported python 3.11 within days after it's official release. But that doesn't mean that the anaconda distribution is also shipping python 3.11 binaries of those libraries, and that is what is needed to make a full anaconda python 3.11 release. If you follow the links @johnnynunez is pointing out, you will see that, to date, there are still no python 3.11 binaries available on anaconda for either numpy or scipy. |
@burnpanck Sorry, you are absolutely right. There are no python 3.11 binaries yet for those packages... |
Due to be released on March 14, fingers crossed! |
Do we have reasons to believe this release date is still accurate? The blog post is from Feb 22 and it also wrongly (I guess) mentions Python 3.10 (rather than 3.11) support in a few places... |
Hello, and thank you for your patience. The 2023.03 Anaconda Distribution installer release will indeed only come for Python 3.8, 3.9 and 3.10. However we did push today most 3.11 packages on our main channel (including numpy and scipy). |
Could you upgrade to 3.11.2 on anaconda channel? The version that exists now it's 3.11 and has some important bugs |
Hello, 3.11.2 is now uploaded to anaconda default channel. |
thanks! |
python 3.11.3 is out |
Python 3.11 support added via conda/conda#12256, this will be included in the May release of conda 23.5.0 |
Checklist
Request type
Name
Python
Version
3.11
Platform(s)
linux-64 (and presumably others)
Location of source code
https://github.com/python/cpython
Conda-Forge feedstock
https://github.com/conda-forge/python-feedstock
Reason for this request
It appears as though 3.10.0 was released (built anyway) a day or so after release on python.org but there doesn't appear to be anything available yet, even though the feedstock updated within a day or so.
(OPTIONAL) Additional information
No response
The text was updated successfully, but these errors were encountered: