Fine-tuning T5 & LLAMA and using RAG

Assalamo Alaikom, I have extensive experience fine-tuning large language models like LLAMA and FLAN T5, as well as implementing Retrieval-Augmented Generation (RAG) solutions during my time at IBM. Based on the video tutorial you shared, I noticed several issues that could be causing the errors you encountered. I can help troubleshoot and resolve these issues efficiently. I propose using my expertise to not only fine-tune the model but also implement the RAG approach with FAISS and vector databases for optimized response retrieval. Additionally, I will compare and evaluate the performance of both methods to ensure the best solution for your medical recommendation model. Given the urgency of your timeline, I will remain available to accelerate the implementation and ensure we meet your project deadline.

Hello rawana49, I am Maryam Abbas, with 4 years of experience in Generative AI. I have carefully read the project requirements and propose to fine-tune T5 & LLAMA models and utilize RAG for response retrieval. By incorporating FAISS for database search and comparing RAG and fine-tuning outputs, I aim to deliver a robust recommendation system, possibly in the medical domain. My portfolio showcases successful AI projects, and I have a track record of positive feedback. Let's discuss your project further to explore the details. Please review my portfolio at https://www.freelancer.pk/u/maryam951 and initiate a chat to proceed. Best regards, Maryam Abbas

If I am not wrong, did you want the recommendation model or question answer based . In one method, you want to use the Hugging based model while in other one , you are interested to use RAG. According to my knowledge, this is one approach . .1. load dataset, 2. split 3. embedding (OpenAi embedding or fake Embedding or pinecode, any other) . 4. vector (Faiss or vector ). then model (model can be pretrained or you can do retraining (finetraining according to your requirements. If I am write, let me dicuss, so we can proceede to the task. thanks.