AttributeError: 'TextGeneration' object has no attribute 'random_state'

[DavidHarar]

Hello,

I followed the amazing tutorial from https://www.youtube.com/watch?v=5a5Dfft-rWc. I ran it smoothly, until I got to the following chunk where we specify our different representations:

from bertopic.representation import KeyBERTInspired, MaximalMarginalRelevance, TextGeneration

# KeyBERT
keybert = KeyBERTInspired()

# MMR
mmr = MaximalMarginalRelevance(diversity=0.3)

# Text generation with Llama 2
llama2 = TextGeneration(generator, prompt=prompt)

# All representation models
representation_model = {
    "KeyBERT": keybert,
    "Llama2": llama2,
    "MMR": mmr,
}
representation_model

The result was

AttributeError: 'TextGeneration' object has no attribute 'random_state'

I looked for it in the code, it looks like there is a random_state parameter that enters as a default to 'TextGeneration', even though it takes no such argument and hence the error arises.

Thanks!

[DavidHarar]

And PS - THANK YOU SO MUCH for the amazing work you did. You're a life-saver.

[MaartenGr]

Thank you for your kind words!

I just recently also ran into this issue. Apparently, scikit-learn had an update that made it necessary to assign parameters to the class itself (which I had completely forgotten!).

However, that should now be fixed in the main branch. So if you install BERTopic from the latest commit, it should work for you.

[DavidHarar]

Perfect, it worked, thanks! Installing from Git directly:

!pip install git+https://github.com/MaartenGr/BERTopic/