fastrank

FastRank Learning to Rank Library written in Rust.


Keywords
coordinate-ascent, learning-to-rank, machine-learning, random-forests, ranklib, rust
License
MIT
Install
pip install fastrank==0.3.2

Documentation

FastRank CI Status Badge PyPI version

My most frequently used learning-to-rank algorithms ported to rust for efficiency.

Read my blog-post announcing the first public version: 0.4. It's alpha because I think the API needs work, not because there's any sort of known correctness or compatiblity issues.

Python Requirements / Release History

  • 0.5 and earlier require only Python 3.5, but no windows builds were pushed.
  • 0.6 requires Python 3.6 due to EOL for Python 3.5 becoming prevalent in the latest pip.
  • 0.6.1 switched to manylinux2010 building; you might get better vectorization from a local copy.
  • 0.7 maintains the requirement of Python 3.6
  • 0.8 and forward will require Python 3.7 so we can use the standard @dataclass annotation and drop the attrs dependency.

Python Usage

pip install fastrank

See this Colab notebook for more, or see a static version here on Github.