Abstract: 

This talk introduces the PyTorch lightning ecosystem one step at a time. We start with Lightning Flash, an AI factory and an easy-to-use framework for getting started with the Lightning ecosystem. With Flash, we showcase a basic NLP task that classifies the toxicity level of user comments. Then, we delve into the core components of the PyTorch Lightning framework. Using these components, we show how to implement state-of-the-art research without worrying about the hassles of engineering.

Session Outline

Part 1: Predefined tasks with Lightning Flash

We showcase a Natural Language Processing (NLP) model for detecting toxicity in plain text with Lightning Flash. We will benefit from utilizing a clean API that allows us to simply plug in any state-of-the-art model from the transformers repository.

Part 2: Intro to PyTorch Lightning core components with Computer Vision

We start with a brief introduction of Pytorch Lightning’s core building blocks and explore how they fit into the typical research/data scientist development flow. We demonstrate the usage of these building blocks while building a standard computer vision task - multi-label classification.

Part 3: Using PyTorch Lightning at Scale

We will then showcase how you can easily scale your training for a large dataset over various accelerators such as GPUs and TPUs. We'll also go through the basic API internals of how PyTorch Lightning succeeds in abstracting the accelerator logic from users with support for distributed strategies, allowing them to focus on writing accelerator-agnostic code.

Background Knowledge

Some familiarity with Python, deep learning terminology, and the basics of neural networks.

Bio: 

Kaushik Bokka is a Senior Research Engineer at Grid.ai and one of the core maintainers of the PyTorch Lightning library. He has prior experience in building production scale Machine Learning and Computer Vision systems for several products ranging from Video Analytics to Fashion AI workflows. He has also been a contributor for few other open source projects and aims to empower the way people and organizations build AI applications.