mklf/word2vec-rs

pure rust implemention of word2vec

Word2Vec-rs

Word2Vec-rs is a fast implemention of word2vec's skip-gram algorithm.

A simple benchmark on a 200M english corpus:

4 threads:

tool	words per sec	memory
word2vec	332k	616M
gensim	311k	249M
word2vec-rs	641K	197M

8 threads:

tool	words per sec	memory
word2vec	611k	616M
gensim	539k	266M
word2vec-rs	995k	203M

• gensim version 0.13.3
• word2vec(google code version): https://github.com/mklf/word2vec

Building

word2vec-rs is written in Rust, so you need a Rust installation in order to compile it( It's super easy). Rust version 1.14 or newer is tested.Building is easy:

git clone https://github.com/mklf/word2vec-rs
cd word2vec-rs
cargo build --release --features=blas
./target/release/word2vec train --help

Running

./target/release/word2vec train input_file_path output_path

word2vec-rs's Issues

Usage example

Hi there,

I myself was wondering whether to implement w2v in rust. I am a python user and noob to rust. In any case, I found your repo and it looks great. Something that could be of great benefit it to add a simple usage example in the README.md

A bit more illustration of what the output looks like? How the input needs be formatted?

Publish to crates.io

Hi i was wondering why dont you publish this crates.io as a crate so everybody can use it without cloning the repo ?