search

pangeo-data / jupyter-earth

Jupyter meets the Earth: combining research use cases in geosciences with technical developments within the Jupyter and Pangeo ecosystems.
https://jupytearth.org
Creative Commons Zero v1.0 Universal
29 stars 6 forks source link

Installing pyMC3 #99

Closed consideRatio closed 2 years ago

consideRatio commented 2 years ago

In #98 i tried to install pyMC3 according to these instructions but failed because...

Encountered problems while solving:
  - package mkl-service-2.3.0-py39h07f9747_1 requires mkl >=2020.2,<2021.0a0, but none of the providers can be installed

conda-forge/noarch        
conda-forge/linux-64      
The command '/bin/sh -c echo "Installing conda packages..."  && mamba install -n ${CONDA_ENV} -y         altair         bqplot         plotly         python-kaleido         seaborn         ipycanvas         ipympl         jupyter_bokeh         jupyterlab-geojson         black         flake8         pep8         pyflakes         pylint         pytest         pytest-cov         jupyter-book         jupytext         numpydoc         sphinx         pymc3         theano-pymc         mkl         mkl-service         ipydatagrid         ipyparallel         lxml         pyhdf         vaex         mhealpy         pytables         statsmodels         xlrd         jupyter-repo2docker         jupyter-vscode-proxy         "jupyterlab>=3.2.4"         "jupyterlab-link-share>=0.2.2"         jupyterlab-git         jupyterlab-system-monitor         nbdime         qgis         retrolab         ipydrawio         websockify         compilers         cxx-compiler         cython         fortran-magic         google-cloud-sdk         sympy         cupy         cudatoolkit=11.2.2  && echo "Installing conda packages complete!"' returned a non-zero code: 1

It isn't clear to me what's the core issue is yet, but I'll try to resolve it and follow up in this issue.

/cc: @abbyazari @facusapienza21 who are interested in this

consideRatio commented 2 years ago

Investigation

Package Source code conda-forge version
mkl IntelPython/mkl v22.0.1
mkl-service IntelPython/mkl-service v2.4.0
pymc3 pymc-devs/pymc v3.11.4
theano-pymc majidaldo/Theano-PyMC v1.1.2

I concluded that the base image pangeo/pangeo-notebook includes nomkl which is designed to prevent any use of mkl, and that led to the observed issue. I opened #100 to remove that and try if we can install pymc3 now. I'm afraid that removing nomkl we end up with other issues, for example a decrease in performance when using numpy or similar.

consideRatio commented 2 years ago

Closing as resolved, import pymc3 now works.

For reference, @fperez made a comment about removing the nomkl package (that is a dummy package to make sure mkl isn't installed) and installing mkl as a dependency: https://github.com/pangeo-data/jupyter-earth/pull/100#issuecomment-1021736812

fperez commented 2 years ago

BTW, it seems to me that this can now be closed, right? In light testing everything worked fine for me - @abbyazari/@facusapienza21, pls let us know if you encounter any issues...