yura

I have fasttext dump trained with python code unsupervised with quantization. When i call getWordVector using java code and python code i see very different vectors.

I noticed that this code actually ignores original scores when reading serialized index. So i was forced to comment it. ``` scala protected GenericTypeahead createTypeahead(GenericTypeaheadConfig config) throws Exception { SOME...

PrefixSelector doesnt properly works with word repetition. I.e. you can type same word(or even short prfix) several times and get same suggestion. You can use something like this to solve...

Training question

7

Am I right saying that this LSTM during training uses its own generated output as input? This could explain low precision of trained models compared to tensorflow implementation. See >...