vhab10/Llama-3-1-8B-Instruct-Unsloth-LoRA-4bit

Overview

This model is a fine-tuned version of the Meta LLaMA 3.1 8B Instruct model. It has been fine-tuned using the Unsloth library with 4-bit quantization for efficient inference and deployment. The fine-tuning process utilized a synthetic dataset from @AI Maker Space consisting of acronyms and their expanded forms in English and was performed using the LoRA (Low-Rank Adaptation) technique, specifically designed for instruction-based tasks. It can be easily deployed in low-resource environments thanks to the 4-bit quantization.

Uploaded model

• Developed by: vhab10
• License: apache-2.0
• Finetuned from model : unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.