This is NOT the LLaMA model released recently converted to work with Transformers. It is NOT that. Simply use this model as you would any other now. Below is an example:
tokenizer = transformers.LLaMATokenizer.from_pretrained("Bitsy/Not-LLaMA-7B-Pytorch-Transformer-Compatible")
model = transformers.LLaMAForCausalLM.from_pretrained("Bitsy/Not-LLaMA-7B-Pytorch-Transformer-Compatible")