Advanced Deep Learning in Pytorch#

Instructor#

  • Instructor: D. Hudson Smith

  • Office: 2105 Barre Hall, Clemson University

  • Email: dane2 AT clemson DOT edu

Workshop Description#

Welcome to the advanced pytorch workshop series. In this series we will start with a simple deep learning example and then iteratively add in more advanced approaches and tooling. We will work through the following techniques:

  • Downloading and fine-tuning pre-trained models

  • Script-based development workflow

  • Organizing model code with Pytorch Ligtning

    • Checkpointing

  • Experiment tracking with Weights & Biases

  • Using multiple-GPUs

    • Using pytorch syntax

    • In Pytorch Lightning

    • Profiling GPU usage

  • Reproducible research with Pytorch

    • Version control

    • Setting random seeds

    • Logging results

  • Hyperparameter tuning

    • Bash scripting

    • Using Sweeps

Prerequisites#

All students should have a Palmetto Cluster account. If you do not already have an account, you can visit our getting started page. Students must have basic familiarity with Python programming and a good grasp of Pytorch fundamentals. This requirement is satisified by taking the the introductory pytorch series or other experience using Pytorch.