No package can be installed for pin: cudatoolkit 12.2.

[pablo-arantes]

Hi,

Until yesterday, I was using CondaColab and everything was working perfectly. Unfortunately, today I encountered the following error:

# >>>>>>>>>>>>>>>>>>>>>> ERROR REPORT <<<<<<<<<<<<<<<<<<<<<<

    Traceback (most recent call last):
      File "/usr/local/lib/python3.10/site-packages/conda/exceptions.py", line 1124, in __call__
        return func(*args, **kwargs)
      File "/usr/local/lib/python3.10/site-packages/mamba/mamba.py", line 941, in exception_converter
        raise e
      File "/usr/local/lib/python3.10/site-packages/mamba/mamba.py", line 934, in exception_converter
        exit_code = _wrapped_main(*args, **kwargs)
      File "/usr/local/lib/python3.10/site-packages/mamba/mamba.py", line 892, in _wrapped_main
        result = do_call(parsed_args, p)
      File "/usr/local/lib/python3.10/site-packages/mamba/mamba.py", line 754, in do_call
        exit_code = install(args, parser, "install")
      File "/usr/local/lib/python3.10/site-packages/mamba/mamba.py", line 547, in install
        solver.add_pin(final_spec)
    RuntimeError: No package can be installed for pin: cudatoolkit 12.2.*

It seems the error is related to Mamba, but I couldn't find any relevant information in their GitHub repository.
Are you experiencing the same issue? Has anyone managed to resolve it?

Thank you,

Pablo

[kimjc95]

Same issue here!
I think the issue is due to the fact that there is no package named 'cudatoolkit' with the version >=12 on conda-forge channel.
Somehow the package name has been changed from 'cudatoolkit' to 'cuda-toolkit' on conda-forge or nvidia channel since the version 12.

[kimjc95]

I came up with a quick workaround.. just erase the pinned file form the conda-meta directory after installing the condacolab.

!rm -rf /usr/local/conda-meta/pinned

[kimjc95]

If you want to use the cuda-toolkit 12.2, you can run this code instead.

with open('/usr/local/conda-meta/pinned', 'w') as f:
    f.write("python 3.10.*\npython_abi 3.10.* *cp310*\ncuda-toolkit 12.2.*")
!mamba install -q -c nvidia cuda-toolkit=12.2 

[jaimergp]

Hi, I added a PR at #63. Nice to have a new CUDA deployed on Friday :P

Let me know if the following installation from branch fixes the issue:

!pip install -U https://github.com/conda-incubator/condacolab/archive/cuda-version-12.tar.gz

import condacolab
condacolab.install()

[jaimergp]

Note that conda-forge only has 12.0 for now, but 12.1 and 12.2 are on the way, so for now I'm only pinning to 12.x (and not 12.2.x).

I guess it should be ok to CUDA compatibility guarantees.

[sef43]

!pip install -U https://github.com/conda-incubator/condacolab/archive/cuda-version-12.tar.gz

This works for the OpenMM Colab notebooks. Thank you!

[raybellwaves]

I recently came across this. I ended up doing !pip install -q condacolab instead of !pip install -q "https://github.com/conda-incubator/condacolab/archive/main.zip". Took me a while to notice commit happen in a the 0.1.x branch https://github.com/conda-incubator/condacolab/tree/0.1.x