Can a dynamic topic model be saved too?

[DominikMann]

Hi folks,

I want to save a GPU-trained model and load it on a CPU-workstation for visualization. The normal models works just fine but I want to export a Dynamic Topic Model too. If I call visualize_topics_over_time() on the CPU workstation, I need to recompute the topics_over_time() model but this takes an awful amount of time on a CPU. Is there a way to save this DTM model like the normal model?
Any help is appreciated!

[MaartenGr]

When you run .topics_over_time the model is not internally saving information with respect to the dynamic topic model but returns all information you need as a dataframe. In other words, when you run the following:

topics_over_time = topic_model.topics_over_time(tweets, timestamps, nr_bins=20)

The information with respect to the dynamic nature of topics is contained in the DataFrame topics_over_time. This means that you would only need to save topics_over_time as a csv locally and pass it later on to .visualize_topics_over_time.

[DominikMann]

Thank you very much for the quick and clear answer! Very appreciated!