Python packages that are non-trivial to install

From ScientificComputing
Revision as of 07:57, 10 July 2024 by Sfux (talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Numpy, scipy

Numpy and scipy are available for all centrally installed Python versions, but if you need the newest version that is available you might need to install it locally in your home directory. Both packages can be linked against the higly-optimized OpenBLAS linear algebra libraries. For this, the path to the OpenBLAS library needs to be specified in an environment variable:

  • Distuils 
OPENBLAS=/cluster/apps/openblas/0.2.13_seq/x86_64/gcc_4.8.2/lib python setup.py install
  • PIP
OPENBLAS=/cluster/apps/openblas/0.2.13_seq/x86_64/gcc_4.8.2/lib python -m pip3 install --user --upgrade numpy
OPENBLAS=/cluster/apps/openblas/0.2.13_seq/x86_64/gcc_4.8.2/lib python -m pip3 install --user --upgrade scipy

Numba/llvmlite

  • module load gcc/4.8.2 python/2.7.6
  • python -m pip3 install --user enum34
  • LLVM_CONFIG=/cluster/apps/llvm/3.6.0/x86_64/bin/llvm-config python -m pip3 install --user llvmlite==0.8.0
  • python -m pip3 install --user numba==0.22.1
Retrieved from "https://scicomp.ethz.ch/w/index.php?title=Python_packages_that_are_non-trivial_to_install&oldid=10057"