Hi, It is not an issue but query .
We want to make word2vec model generation incremental. Currently, it takes ~10 hrs to build a model. Everyday new data comes and we don't want to build new model from scratch ... instead want to update it with new data.
Is it possible in gensim ??
Yes. You can use the train method to continue training.
This limits vocabulary to what you initially passed to build_vocab though. If your vocabulary is static (known beforehand), this is not a problem. If you need to be adding new words dynamically as well though, check out @rutum's pull request #365 .
That was quick ... thanks
Most helpful comment
Yes. You can use the
trainmethod to continue training.This limits vocabulary to what you initially passed to
build_vocabthough. If your vocabulary is static (known beforehand), this is not a problem. If you need to be adding new words dynamically as well though, check out @rutum's pull request #365 .