LoRA extraction from Gradient AI's https://huggingface.co./gradientai/Llama-3-8B-Instruct-Gradient-1048k model.
LoRA extraction only targeted from the self_attn modules.
Rank: 1024