simple-fasttext-server

You send an http get with some text, and get label probabilities for it.

Installation

1. install leiningen if not already installed
2. clone this repo
3. change directory to the included directory fasttext.
4. place your trained fasttext model here ― it should be named classifier.bin.
5. git clone https://github.com/facebookresearch/fastText.
6. validate the installation by running lein test

Usage

to start the server:

lein run [port]

to get a response, make an http get request with the single query parameter text. E.g. with curl this would look like ―

curl -G http://localhost:3001/predict --data-urlencode "text=classify me"

Performance notes

Currently utilizes just one fasttext worker! support for multiple workers is not built-in yet. See below the VisaulVM output of the performance and throughput tests (included in the test directory) as of the current commit on a 12-core i7 ―