Here, specific methods to modify and improve the pre-trained Meta Llama 3 will be detailed. In specific, steps needed to modify the Llama 3 open source code architecture for re-training will be explained. In addition, advanced methods to fine-tune Llama 3 using PEFT/LoRA techniques to reduce time and parameter size will be covered. RAG (retrieval augmented generation) to improve accuracy and source citation capabilities of live data as a Generative AI Assistant will be highlighted. Lastly, current built-in Llama 3 agentic behavior and future Drug Discovery use will be illustrated.

Reference: AI Engineer. (2023b, November 15). Building Production-Ready RAG Applications: Jerry Liu [Video]. YouTube.    • Building Production-Ready RAG Applica...  
Reference: ChemicalQDevice - Meta Llama 3 Fine-tuning, RAG. (n.d.). https://www.chemicalqdevice.com/meta-...