Llama 2 is a state-of-the-art large language model (LLM) developed by Meta AI, designed for a variety of natural language processing (NLP) tasks. Its architecture builds upon transformer-based models, leveraging massive text corpora during pretraining to develop rich language understanding capabilities. Fine-tuning Llama 2 can be customized to a specific task by using a smaller, task-specific dataset, often resulting in a specialized model that outperforms the general-purpose base model on that task. In this project, we fine-tune the Llama-2-7b-hf model using a subset of the Guanaco dataset, focusing on developing a highly efficient model called "MiniGuanaco."

This repository demonstrates the steps and code required to fine-tune Llama 2 for specific tasks. Using the Hugging Face model NousResearch/Llama-2-7b-hf as the base, the model is fine-tuned with the dataset mlabonne/guanaco-llama2-1k. The fine-tuned model is saved under the name llama2-miniguanaco.