/EECE-7398-Machine-Learning-with-Small-Data---Fall-2024

EECE 7398: Machine Learning with Small Data - Fall 2024

Primary LanguageJupyter Notebook

EECE 7398: Machine Learning with Small Data - Fall 2024

Leveraging the Discovery Cluster for Advanced ML Research

Instructor: Prof. Sarah Ostadabbas

Northeastern University, Department of Electrical and Computer Engineering

Overview

This repository contains the materials for the EECE 7398: Machine Learning with Small Data course. The course emphasizes using Northeastern's Discovery Cluster for advanced machine learning experiments, particularly in small data scenarios.

The materials include pre-prepared slides (uploaded directly as .pdf) and hands-on exercises to guide students through accessing, setting up, and utilizing the Discovery Cluster for machine learning research. Topics covered:

  • Introduction to the Discovery Cluster
  • Cluster access and environment configuration
  • PyTorch setup and GPU utilization
  • Monitoring experiments with Weights & Biases
  • Best practices for managing machine learning experiments

Contents

  • Slides/: Uploaded PDF slides covering session topics
  • cifar10_classification_assignment_corrected.ipynb: CIFAR-10 classification assignment with W&B monitoring
  • cifar10_training_inference_part3.ipynb: Notebook for training/inference on the Discovery Cluster
  • wandb_monitoring_script.py: Script for setting up W&B monitoring
  • conda_environment_setup_part1.ipynb: Notebook for setting up the Conda environment
  • README.md: This readme file

Prerequisites

Before using the Discovery Cluster for this course, students must have:

  1. Access to the Discovery Cluster: Request access through ServiceNow
  2. Basic knowledge of Python and machine learning
  3. Familiarity with Conda for environment management

Exercises Overview

  1. CIFAR-10 Classification and W&B Monitoring (cifar10_classification_assignment_corrected.ipynb)

    • Train and track a custom CNN on the CIFAR-10 dataset.
    • Utilize Weights & Biases (W&B) to monitor training performance.

  2. Training and Inference on GPUs (cifar10_training_inference_part3.ipynb)

    • Train a simple classifier using the CIFAR-10 dataset on the Discovery Cluster GPUs.
    • Run inference on pre-trained models.

  3. W&B Monitoring Setup (wandb_monitoring_script.py)

    • A Python script to set up W&B for experiment tracking without Jupyter Notebooks.

  4. Setting up Conda Environment (conda_environment_setup_part1.ipynb)

    • Step-by-step guide to creating a Conda environment with the required dependencies for GPU-accelerated PyTorch on the Discovery Cluster.

How to Use

  1. Clone the repository:

    git clone https://github.com/your-repo/EECE7398-ML-Small-Data-Fall2024.git

  2. Review the provided materials:

    • Use the uploaded slides in Slides/ to follow along with the course sessions.
    • Work through the exercises in the provided Jupyter notebooks.

  3. Access the cluster and follow the steps outlined in the slides to set up your environment for machine learning experimentation.

Additional Resources

For any issues or questions, feel free to reach out to the Northeastern RC support team via ServiceNow.

License

This repository is licensed under the MIT License. See the LICENSE file for more information.