Hacker News Reader

NB: The libraries use Torch.

I'm still playing around with fasttext (which is amazing btw) officially announced last week so I'm surprised to see Facebook Research announce and release another project so soon.

View on HN

Could you share what you did with fasttext? It'd be great to see some examples around the amazing tools being open sourced! I read that there are no documentation comments, maybe they were removed?

For basic use, you can construct text vectors faster/better than word2vec/doc2vec.

The Python package which serves as an API to fasttext (https://pypi.python.org/pypi/fasttext/0.7.2 ) is well documented, is constantly being updated, and is easy to use.

Does/can fastText use GPGPU at all? I use a CUDA cbow word2vec [1] that is really quick, and has served me quite well. Hard to know if switching to something more accurate would be worth giving up the speed :)

[1] https://github.com/ChenglongChen/word2vec_cbow

No, fastText is CPU only.

fastText is so fast, you may not need to use the GPU. That was their goal from the very beginning.

Reply on news.ycombinator.com