'UnsatisfiableError glibc and cudatoolkit
conda install -c anaconda -c conda-forge -c nvidia cudatoolkit=11 tensorflow-gpu=2
I want to install cudatoolkit=11 and tensorflow-gpu=2, but conda gave me the following error message
UnsatisfiableError: The following specifications were found to be incompatible with your system:
- feature:/linux-64::__glibc==2.31=0
- cudatoolkit=11 -> __glibc[version='>=2.17,<3.0.a0']
Your installed version is: 2.31
If I run
conda install -c anaconda -c conda-forge -c nvidia cudatoolkit=11
conda can successfully install cudatoolkit=11.
Since glibc is a system feature, why installing cudatoolkit by itself is ok?
Why isn't glic 2.31 in range >=2.17,<3.0.a0?
Solution 1:[1]
I am providing answer as I could not find outright reason and solution.
If you are installing from conda-forge channel in conda base env this problem is very likely to occur due to c-runtime difference between conda and conda-forge.
To solve this , create virtual-env make conda-forge channel as first and stick to it(Strict) for installing packages, I don't see any harm in sticking to conda-forge , and have found many experienced people suggesting same (stick to conda-forge channel).
Solution 2:[2]
Updating anaconda to the newest version solved my problem.
Try conda update conda?
Solution 3:[3]
Install fvcore and iopath with conda install -c fvcore -c iopath -c conda-forge fvcore iopath. Then try again.
Solution 4:[4]
Let the environment solver choose the correct versions for cuda and tensorflow.
assuming you are setting up a new conda environment:
conda create -n <NAME> cudatoolkit tensorflow-gpu <packages> -c <channels>
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|---|
| Solution 1 | Deep Saurabh |
| Solution 2 | LeoX |
| Solution 3 | Michael |
| Solution 4 |