Google colab and CUDA 12.2

[Ash100]

Hello Ryan,
Thanks for writing such great notebooks for the community. Google colab is shifted to CUDA 12.2. As a result, the mamba installation script are not getting success.
!mamba install -c conda-forge -c bioconda mgltools=1.5.7 biopython=1.80 openbabel=3.1.1 plip=2.2.2 zlib=1.2.13 xlsxwriter=3.0.3

The installation log generates,

No package can be installed for pin: cudatoolkit 12.2.*

I tried different things, like mentioning cudatoolkit=12.2 in the script, also remove the package version details, so that conda installs the appropriate version by itself. Nothing is working.
Will you be able to update the code?
Thanks

[RyanZR]

Hi @Ash100, I will be looking into it soon.
Thank you for the heads up.

[RyanZR]

This issue is reported in issue #60. The presence of CUDA v12.2 causes condacolab to inject a pin of cudatoolkit 12.2.*. Since Conda Forge isn't yet building cudatoolkit=12.2, this breaks Mamba for now.

Two quick temporary workarounds would be:
1. Remove the pinned file from Conda-Meta

!pip install condacolab==0.1.7
import condacolab
condacolab.install_mambaforge()
!rm -rf /usr/local/conda-meta/pinned # ADD THIS LINE HERE
!mamba env update --file environment.yml

2. Install archived CondaColab with CUDA 12.x support

!pip install -U https://github.com/conda-incubator/condacolab/archive/cuda-version-12.tar.gz # REPLACE WITH THIS LINE
import condacolab
condacolab.install_mambaforge()
!mamba env update --file environment.yml

Until PR #63 gets merged in, applying either one of the workaround should fix the issue for now. Let me know if this solve the problem.

[RyanZR]

Hi @Ash100, PR #63 has been merged. The installation process should be working fine again.
Please try the notebooks at the develop branch.