Adding `safetensors` variant of this model

Llama 2 5 epochs trained on custom legal dataset

initial commit