recover topics from saved model

[aminathshausan]

Hi,
I am new to using BerTopic. I have loaded a previously saved model. How do I get the topics from the loaded model to perform the dynamic visualization in this step:
topics_over_time = topic_model.topics_over_time(abstracts, topics, years)

Thanks

[MaartenGr]

It depends on the version of BERTopic you are using. If you are using v0.12, then you would not need to pass topics at all. However, if you are using an earlier version, I believe it would suffice to get those topics by running

topics = topic_model._map_predictions(topic_model.hdbscan_model.labels_)

[aminathshausan]

Thanks. I did uninstall bertopic, reinstalled the latest version (0.12.0) and tried to import BERTopic using
from bertopic import BERTopic
I am now getting the following Type error:
TypeError: __init__() got an unexpected keyword argument 'cachedir'
What does this mean and how do I fix it please.

I am getting the same Type error when installing version 0.11.0 too
Thanks

[MaartenGr]

This is a known issue and you can find more information about that here and here.

[aminathshausan]

Thanks, I could fix the error following the links, and successfully imported BERTopic. But loading the saved model now gives the following error with BERTopic version 0.11.0.

AttributeError: Can't get attribute '_rebuild_function' on <module 'numba.core.serialize' from '/Users/.../opt/anaconda3/lib/python3.8/site-packages/numba/core/serialize.py'>

Any help please
Thanks

[MaartenGr]

Did you make sure that the environment in which you saved the model is exactly the same as the environment in which you loaded the model? Seeing as you did not have this issue saving the model, most likely something has changed between environments. I would advise starting from a completely fresh environment, making sure you keep track of dependencies and re-train your model. That way, by tracking the package versions in your environment, you prevent most of these issues.

[aminathshausan]

Thanks- that worked