Multiple per-topic-word-priors crash jupyter kernel

bab2min / tomotopy

Python package of Tomoto, the Topic Modeling Tool

https://bab2min.github.io/tomotopy

MIT License

548 stars 62 forks source link

Multiple per-topic-word-priors crash jupyter kernel #125

Closed ebergam closed 3 years ago

ebergam commented 3 years ago

Hi @bab2min ! I was trying to do something similar the guidedLDA model (which I think would be worth adding to tomotopy, it is very powerful!).

I tried to add more then one word prior, with same probability, for several topics, using your example code.

As soon as I try to train the model, the kernel dies. I tested the same on two different machines (linux and windows), and I have uploaded data+code here for you to replicate. Am I making a conceptual mistake, or there is a bug? Thank you!

bab2min commented 3 years ago

Hi @ebergam Thank you for reporting the bug. I'll examine it.

bab2min commented 3 years ago

Hi @ebergam, It has been finally proved that LDAModel.set_word_prior causes a crash after add_doc called. I will update the patch that fixes this bug soon.

Thank you again for reporting the bug.