A way to save trained word vectors?

[park]

Nice work!

Is there a way to save the trained word vectors into text or binary files for later use?

Thanks.

[EddieOne]

checkpoints are one way, https://www.tensorflow.org/programmers_guide/variables#checkpoint_files

[MrRexZ]

I think given that the tensorflow graph node implementation has shape that is reliant on the vocab size, evaluating it into a numpy array and saving the parameters into a pickle file would be more effficient. Otherwise if by using checkpoint, one would have to load the model with old graph with the nodes shape of old vocab size, recreate the graph with the new vocab size after fitting the model to corpus.

[Seafoodair]

nice work

[EddieOne]

I forgot to post it, I did manage this. https://github.com/EddieOne/tf-glove-deluxe