* Training and other expensive methods are interruptible (e.g. by typing Ctrl-C) * Better const correctness * Update fastText source code (add subwords support in supervised models)
0.3.0
* Support older fastText models (before quantization support) * Remove useless parameter * Update fastText source code to latest version
0.2.2
* Add documentation for model.words_for_vector() (thanks yaakov2) * Add continuous integration (build and tests) using Travis CI * Change variant library in order to support g++ 4.7 and 4.8
0.2.1
Add a new `model.words_for_vector(vector, k)` method. (Thanks yaakov2)
0.2.0
* Fix some bugs * Add support for numpy ndarray (optional) * Update the fastText source to the latest version
0.1.0
Initial release!
* Fairly complete support of all the features of the fastText binary (training, prediction, quentization) * Add some new features too (normalized probabilities, label extraction, ...)