In the code for action recognition, I'm a bit confused about the part where pre-trained weights for the LLaMA model are imported. Can you please point out the section of code that handles the import of pre-trained weights for the LLaMA model?
Can you provide the pre-trained weights for the LLaMA model you are using?
This line loads the weights of LLaMA. I wrote a function in llama.py that supports loading one layer of transformer at here.
Sorry about it that I am not sure if I am allowed to distribute the weights of LLaMA publicly. But I just made some updates to the doc of action recognition explaining the preparation of LLaMA. Hope this helps!