{"repo":"mirzayasirabdullahbaig07/Fine-Tuning-LLaMA-3.2-3B-Using-PEFT-LoRA","free":true,"listed":false,"github":"https://github.com/mirzayasirabdullahbaig07/Fine-Tuning-LLaMA-3.2-3B-Using-PEFT-LoRA","clone":"git clone https://github.com/mirzayasirabdullahbaig07/Fine-Tuning-LLaMA-3.2-3B-Using-PEFT-LoRA.git","description":"This project showcases parameter-efficient fine-tuning of the LLaMA 3.2 (3B) language model using PEFT (Parameter-Efficient Fine-Tuning) and LoRA (Low-Rank Adaptation). It is optimized for minimal resource usage and trained on a domain-specific dataset to enhance performance in specialized tasks.","language":"Jupyter Notebook","stars":46,"topics":["artificial-intelligence","data-visualization","deep-learning","llama","machine-learning","python"],"license":null,"category":"machine-learning","readme_excerpt":"🧠 Parameter-Efficient Supervised Fine-Tuning of LLaMA 3.2 (3B) on a Medical Chain-of-Thought Dataset 👨‍⚕️ Project Objective This project fine-tunes the LLaMA 3.2 (3B) language model on a Medical Chain-of-Thought (CoT) dataset using parameter-efficient fine-tuning (PEFT) with LoRA via the Unsloth library on Kaggle Notebooks . The goal is to enhance the model’s capability to reason and generate structured responses in the medical domain. 🧪 Key Features - Utilizes LoRA (Low-Rank Adaptation) for efficient fine-tuning. - Trains on the FreedomIntelligence/Medical-CoT dataset from Hugging Face. - Applies structured formatting using and tags. - Tracks training progress and metrics using Weights & Biases (wandb) . - Evaluates model improvements using ROUGE-L scores. - Uploads the fine-tuned adapter and tokenizer to Hugging Face Hub . - Provides clear instructions for inference using the adapter. 🔧 Workflow Overview 1. Environment Setup - Kaggle Notebook configured with required libraries. - GPU enabled for faster computation. - wandb initialized for live training logs. 2. Dataset Preparation - Loads Medical Chain-of-Thought dataset. - Formats data using and tags. - Splits into training and validation subsets. 3. Model Loading - Loads the quantized 4-bit LLaMA 3.2 3B base model via Unsloth. - Applies LoRA using PEFT for efficient training. 4. Data Preprocessing - Tokenizes and pads inputs. - Prepares datasets for supervised fine-tuning. 5. Training - Trains the LoRA adapter using t","default_branch":null,"files":null,"tree":[],"storefront":"/r/mirzayasirabdullahbaig07","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/mirzayasirabdullahbaig07/Fine-Tuning-LLaMA-3.2-3B-Using-PEFT-LoRA/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}