Rankings_watch displaying wrong sequence

jalammar / ecco

Explain, analyze, and visualize NLP language models. Ecco creates interactive visualizations directly in Jupyter notebooks explaining the behavior of Transformer-based language models (like GPT2, BERT, RoBERTA, T5, and T0).

https://ecco.readthedocs.io

BSD 3-Clause "New" or "Revised" License

1.98k stars 169 forks source link

Rankings_watch displaying wrong sequence #84

Open MiriUll opened 2 years ago

MiriUll commented 2 years ago

Hello, I have a problem with the rankings_watch() function. I used a predefined GPT2 model and gave it the input "Today, the weather is". However, in the visualization, only the first token is shown although the model creates the output correctly:

Thank you for your help :D

jalammar commented 2 years ago

Yep, I can replicate this. It's a an issue with the title of the figure. But the values are correct for the 5th position.

Notice how "getting" is number 1 and is what the model actually outputs.