Closed leofang closed 4 years ago
ref: conda-forge/conda-forge-pinning-feedstock#352
@jakirkham It seems our custom pinning didn't take effect? We built and pinned at 7.6.4, but I'm seeing this (7.6.5 was just out last week: https://github.com/AnacondaRecipes/cudatoolkit-feedstock/pull/2#issuecomment-567693940):
$ conda install cupy
Collecting package metadata: done
Solving environment: done
## Package Plan ##
environment location: /home/leofang/conda_envs/from_CF
added / updated specs:
- cupy
The following packages will be downloaded:
package | build
---------------------------|-----------------
_libgcc_mutex-0.1 | main 2 KB https://conda.anaconda.org/conda-forge
ca-certificates-2019.11.28 | hecc5488_0 145 KB https://conda.anaconda.org/conda-forge
certifi-2019.11.28 | py37_0 148 KB https://conda.anaconda.org/conda-forge
cudnn-7.6.5 | cuda10.1_0 250.6 MB https://conda.anaconda.org/anaconda
cupy-7.0.0 | py37ha7c4746_0 23.9 MB https://conda.anaconda.org/conda-forge
libgcc-ng-9.2.0 | hdf63c60_0 8.6 MB https://conda.anaconda.org/conda-forge
libstdcxx-ng-9.2.0 | hdf63c60_0 4.5 MB https://conda.anaconda.org/conda-forge
nccl-2.5.6.1 | h51cf6c1_0 96.8 MB https://conda.anaconda.org/conda-forge
------------------------------------------------------------
Total: 384.6 MB
The following NEW packages will be INSTALLED:
_libgcc_mutex conda-forge/linux-64::_libgcc_mutex-0.1-main
cudatoolkit anaconda/linux-64::cudatoolkit-10.1.243-h6bb024c_0
cudnn anaconda/linux-64::cudnn-7.6.5-cuda10.1_0
cupy conda-forge/linux-64::cupy-7.0.0-py37ha7c4746_0
nccl conda-forge/linux-64::nccl-2.5.6.1-h51cf6c1_0
The following packages will be UPDATED:
ca-certificates 2019.9.11-hecc5488_0 --> 2019.11.28-hecc5488_0
certifi 2019.9.11-py37_0 --> 2019.11.28-py37_0
libgcc-ng anaconda::libgcc-ng-9.1.0-hdf63c60_0 --> conda-forge::libgcc-ng-9.2.0-hdf63c60_0
libstdcxx-ng anaconda::libstdcxx-ng-9.1.0-hdf63c60~ --> conda-forge::libstdcxx-ng-9.2.0-hdf63c60_0
The explicit pinning (via
run_export
) https://github.com/conda-forge/cupy-feedstock/blob/eddbda9367c972757552646d1abe8ac8c1662e44/recipe/meta.yaml#L45 can be avoided once the upstream (conda-forge-pinning) is updated.