UnsatisfiableError when installing cuml

Question

I am installing cuml using conda

conda install -c rapidsai cuml=21.12

This is the trace I get:

Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: /
Found conflicts! Looking for incompatible packages.
This can take several minutes.  Press CTRL-C to abort.
failed

UnsatisfiableError: The following specifications were found to be incompatible with each other:

Output in format: Requested package -> Available versionsThe following specifications were found to be incompatible with your system:

  - feature:/linux-64::__glibc==2.31=0
  - cuml=21.12 -> cupy[version='>=7.8.0,<10.0.0a0'] -> __glibc[version='>=2.17|>=2.17,<3.0.a0']
  - python=3.8 -> libgcc-ng[version='>=7.5.0'] -> __glibc[version='>=2.17']

Your installed version is: 2.31

My confusion is that it's saying I need a __glibc version of greater than 2.17 but I have an installed version of 2.31.

Answer 1

Ignoring Conda's poor conflict reporting, RAPIDS has very specific channel specification (rapidsai > nvidia > conda-forge), so that could be affecting the solving. It might be sufficient to include those channels:

conda install -c rapidsai -c nvidia -c conda-forge cuml=21.12

However, it could also be the case that something previously installed is preventing correct installation. Moreover, when packages require specialized channels it is generally better practice to create a dedicated environment:

conda create -n cuml -c rapidsai -c nvidia -c conda-forge cuml=21.12

This latter is similar to what the RAPIDS installation selector generates.

Answer 2

Besides trying, this

conda create -n cuml -c rapidsai -c nvidia -c conda-forge cuml=21.12

from the previous post of @merv, just try that:

conda create -n cuml -c rapidsai -c nvidia -c conda-forge cuml

should solve the problem.