Open flomock opened 4 years ago
You look to have both opt-einsum
and opt_einsum
installed with different versions. The former is coming from conda-forge. If you create an environment using only packages from the anaconda
or defaults
channels does this problem persist?
Seams that only opt_einsum is installed, but then I get the error:
pkg_resources.ContextualVersionConflict: (scipy 1.5.0 (/home/go96bix/programs/conda/envs/vidhop_test/lib/python3.7/site-packages), Requirement.parse('scipy==1.4.1; python_version >= "3"'), {'tensorflow'})
seems like one of my packages expects scipy 1.4.1 but that requirement has not been part of the environment requirements during installation. But I'm not sure where this requirement is coming from, maybe it's not from the tensorflow package.
conda list --show-channel-urls
Actual Behavior
pkg_resources.ContextualVersionConflict: (opt-einsum 0+untagged.56.g2664021.dirty (/home/go96bix/programs/conda/envs/vidhop_test/lib/python3.7/site-packages), Requirement.parse('opt-einsum>=2.3.2'), {'tensorflow'})
tensorflow 2.2.0 expects opt-einsum>=2.3.2 But the real name of the package is opt_einsum . There is a version falsely named 3.0.0, but this version only works with scipy 1.4.1 but your package installs version 1.5.Expected Behavior
Fix the naming, I think then the environment solving works as expected.
Steps to Reproduce
conda install -c tensorflow
Anaconda or Miniconda version:
Operating System:
conda info
conda list --show-channel-urls