TCC Poli USP Pro

FineTuning LLMs on conversational medical dataset.

The lexical simplification baseline for MLSP Shared Task at BEA 2024

Repositório dedicado ao trabalho de graduação

🦙 Llama 2 7B fine-tuned to revive Rick

A cybersecurity chatbot built using open-source LLMs namely Falcon-7B and Llama-2-7b-chat-hf. The models are fine-tuned on a custom question-answer dataset compiled from the OWASP Top 10 and CVEs from NVD.

Developed and fine-tuned prompts to optimize LLaMA-2's performance for tasks like text generation, document analysis, and chatbot creation. Utilized iterative prompt refinement, few-shot learning, and context-driven chatbot behavior to demonstrate practical applications of LLaMA-2.

LLM-powered Chatbot for arXiv Knowledge Retrieval

This project demonstrates how to fine-tune a large language model (LLM) for the IT domain using Amazon SageMaker, creating an "IT Domain Expert" model.

AI-powered email management system that automates email categorization, sentiment analysis, summarization, and response generation using NLP models and machine learning, enhancing communication efficiency.

AI-based Psychology Assistant built with llama-2

Retrieval-Augmented Generation (RAG) to analyze and respond to queries about President Biden's 2023 State of the Union (SOTU) Address

Quantization of the Llama 2 model

This repository contains the supplementary material / appendix to go with the paper “Is Temperature the Creativity for Large Language Models” by Max Peeperkorn, Tom Kouwenhoven, Dan Brown, and Anna Jordanous.

Docker image for LLaVA: Large Language and Vision Assistant

Discussed about 4 use-cases or case studies. Discussed about the approaches and significance of these use-cases as these are different from others. There are several approaches available which can be done using LLM but here the approaches and it's significance could bring insightful approaches towards it's execution.

The common setup to run LLM locally. Use llama-cpp to quantize model, Langchain for setup model, prompts, RAG, and Gradio for UI.

Fine-tuning LLaMA 2 for toxicity classification using a balanced Kaggle dataset, with a focus on overcoming class imbalance, optimizing computational efficiency through PEFT and QLORA, and achieving high accuracy in detecting toxic content across multiple classes

This project aims to develop an API backend using Flask to support Text Generation Inference with Llama v2 Models

A Streamlit app for document question answering and text summarization.