search

VarIr / scikit-hubness

A Python package for hubness analysis and high-dimensional data mining
BSD 3-Clause "New" or "Revised" License
44 stars 9 forks source link

v0.30: Documentation #92

Open VarIr opened 2 years ago

VarIr commented 2 years ago

All documentation must be adapted to the new API, including example use-cases.

VarIr commented 2 years ago

Quickstart example of hubness estimation, hubness reduction and usage in sklearn workflows updated in the readme in #105. Userguide and further examples upcoming.

jolespin commented 5 months ago

Is there a beta release for v0.30?

VarIr commented 5 months ago

The main branch here is at v0.30. I just tested the following, which works for me:

conda create --prefix /tmp/scikit-hubness python=3.10
conda activate /tmp/scikit-hubness
pip install https://github.com/VarIr/scikit-hubness/archive/main.tar.gz 
python3
# ... and running the quickstart example code from README.md
jolespin commented 5 months ago

Awesome, thanks for your help. This works for me too w/ v0.30a.

Apologies for the mixup. If it helps, I was able to reproduce the error from last time which happened both on Apple Silicon and Linux server I have access to. 

mamba create -y -n test_env -c conda-forge python=3.10
conda activate test_env
pip install scikit-hubness

Error looks like this: 

      OSError: [Errno 66] Directory not empty: '/private/var/folders/kx/9xfkrf_10pj5kdyh9my45twm0000gn/T/pip-install-l7ocvwkq/nmslib_2bcd54c85a904f4b9c86c5b930bdbdd7/.eggs/numpy-1.26.4-py3.10-macosx-10.9-x86_64.egg/numpy-1.26.4.dist-info' -> '/private/var/folders/kx/9xfkrf_10pj5kdyh9my45twm0000gn/T/pip-install-l7ocvwkq/nmslib_2bcd54c85a904f4b9c86c5b930bdbdd7/.eggs/numpy-1.26.4-py3.10-macosx-10.9-x86_64.egg/EGG-INFO'
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed cleaning build dir for nmslib
Successfully built annoy falconn
Failed to build scikit-learn nmslib
ERROR: Could not build wheels for scikit-learn, nmslib, which is required to install pyproject.toml-based projects
VarIr commented 5 months ago

Thank you. So building some dependencies fails when installing an old version from PyPI. This is not entirely surprising in particular for nmslib, which hasn't been maintained since 2021 or so.