search

jessevig / bertviz

BertViz: Visualize Attention in NLP Models (BERT, GPT2, BART, etc.)
https://towardsdatascience.com/deconstructing-bert-part-2-visualizing-the-inner-workings-of-attention-60a16d86b5c1
Apache License 2.0
6.84k stars 773 forks source link

Request for adding the transformers_neuron_view for LLAMA series models #130

Open iBibek opened 6 months ago

iBibek commented 6 months ago

Hi,

This repository has been really helpful. I would like to request adding transformers_neuron_view for the LLAMA series models as well.

Thank you.

Bachstelze commented 4 months ago

Are the key, value and query matrix relevant, if the output is every time the same for instruction-tuned GPTs? https://github.com/jessevig/bertviz/issues/128#issuecomment-1914344087