adjidieng / ETM

Topic Modeling in Embedding Spaces
MIT License
538 stars 126 forks source link

Fix problem with training embeddings from scratch #31

Closed ahoho closed 3 years ago

ahoho commented 3 years ago

The current sentence iterator doesn't close the file, which I think (not sure) causes training to terminate prematurely. I was getting very low quality embeddings and short training times, but after this fix, embeddings become sensible again.

Thanks very much for putting up the code!

adjidieng commented 3 years ago

Thank you!