PyTorch nightly cannot be installed at the same time as numpy in a conda env on osx-64 / Python 3.11 #97031
Description
CONDA_SUBDIR=osx-64 \
conda create -n tv311 --dry-run \
-c pytorch-nightly -c defaults \
python=3.11 pytorch numpy'>=1.11'(You can drop the CONDA_SUBDIR env var if you already on this system)
Collecting package metadata (current_repodata.json): done
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: -
Found conflicts! Looking for incompatible packages. failed
UnsatisfiableError: The following specifications were found to be incompatible with each other:
Output in format: Requested package -> Available versions
Package python conflicts for:
python=3.11
pytorch -> python[version='>=2.7,<2.8.0a0|>=3.10,<3.11.0a0|>=3.11,<3.12.0a0|>=3.8,<3.9.0a0|>=3.9,<3.10.0a0|>=3.7,<3.8.0a0|>=3.6,<3.7.0a0']
pytorch -> filelock -> python[version='>=3.5|>=3.5,<3.6.0a0|>=3.6|>=3.7|>=3.8']
Package numpy conflicts for:
pytorch -> networkx -> numpy[version='>=1.19']
pytorch -> numpy[version='>=1.11.3,<2.0a0|>=1.16.6,<2.0a0|>=1.19,<2|>=1.21,<2|>=1.19.2,<2|>=1.21.2,<2|>=1.21.2,<2.0a0|>=1.14.6,<2.0a0']
numpy[version='>=1.11']
Package setuptools conflicts for:
python=3.11 -> pip -> setuptools
pytorch -> jinja2 -> setuptools
Same environment solves fine for other platforms and even for osx-arm64.
conda being conda, the shown conflicts are actually all fine. Trying again with manually resolved versions:
CONDA_SUBDIR=osx-64 \
conda create -n tv311 --dry-run \
-c pytorch-nightly -c defaults \
python=3.11 pytorch==2.1.0.dev20230317 numpy==1.23.5
gives
Collecting package metadata (current_repodata.json): done
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: \
Found conflicts! Looking for incompatible packages. failed
UnsatisfiableError: The following specifications were found to be incompatible with each other:
Output in format: Requested package -> Available versions
Package python conflicts for:
pytorch==2.1.0.dev20230317 -> python[version='>=3.10,<3.11.0a0|>=3.11,<3.12.0a0|>=3.8,<3.9.0a0|>=3.9,<3.10.0a0']
numpy==1.23.5 -> python[version='>=3.10,<3.11.0a0|>=3.11,<3.12.0a0|>=3.9,<3.10.0a0|>=3.8,<3.9.0a0']
python=3.11
numpy==1.23.5 -> mkl-service[version='>=2.3.0,<3.0a0'] -> python[version='>=2.7,<2.8.0a0|>=3.6,<3.7.0a0|>=3.7,<3.8.0a0|>=3.5,<3.6.0a0']
pytorch==2.1.0.dev20230317 -> filelock -> python[version='>=2.7,<2.8.0a0|>=3.6,<3.7.0a0|>=3.6|>=3.7|>=3.7,<3.8.0a0|>=3.5,<3.6.0a0|>=3.8|>=3.5']
Package numpy conflicts for:
pytorch==2.1.0.dev20230317 -> networkx -> numpy[version='>=1.19']
numpy==1.23.5
Package setuptools conflicts for:
pytorch==2.1.0.dev20230317 -> jinja2 -> setuptools
python=3.11 -> pip -> setuptools
Again, there is no conflict there. Using --solver=libmamba to get more sensible conflicts reveals:
- package numpy-1.23.5-py311hff9dab5_0 requires blas * openblas, but none of the providers can be installed
Not sure if that is helpful either, since
$ CONDA_SUBDIR=osx-64 conda search blas
Loading channels: done
# Name Version Build Channel
blas 1.0 mkl pkgs/main
blas 1.0 openblas pkgs/main
🤷