Lab 7: Web Deployment
In this lab, we'll take the leap from ML model to ML-powered application by packaging our text recognizer into a portable TorchSript binary, wrapping that binary up into a serverless cloud function, and building a simple UI in Python with gradio.
Outline
- 00:00 Overview
- 01:06 Compiling the model to TorchScript
- 06:00 Why not deploy on GPUs?
- 08:58 Building a GUI with gradio
- 15:34 Spinning up a model service
- 21:11 Creating a public URL with ngrok
- 24:52 Writing a Dockerfile for our server
- 30:06 Recap