Closed ynqa closed 7 years ago
Is the reason for this smaller binary files?
Simply set the format the same as original word2vec output. I also thought the indexes of word vector are unnecessary info.
This way you can just use Gorgonia's serialization methods and dump them
Overview
<word> <index1>:<value1> <index2>:<value2> ...
<word> <value1> <value2> ...