{"id":5647027,"date":"2022-12-12T17:01:31","date_gmt":"2022-12-12T22:01:31","guid":{"rendered":"https:\/\/lightning.ai\/pages\/?p=5647027"},"modified":"2023-03-07T17:17:39","modified_gmt":"2023-03-07T22:17:39","slug":"ml-training-deployment","status":"publish","type":"post","link":"https:\/\/lightning.ai\/pages\/community\/tutorial\/ml-training-deployment\/","title":{"rendered":"How to Build a Machine Learning Training and Deployment Pipeline"},"content":{"rendered":"

Key Takeaways<\/h3> Learn how you can use Lightning to build a model training and deployment pipeline that is customizable, integrated with t<\/span>ools (like monitoring, data warehouses, and feature stores), and deployed in production. <\/div>\n

MLOps is essential for companies both large and small that build products and services powered by AI. Given the wide variety of tools and platforms that aim to solve different parts of the machine learning lifecycle, choosing between them isn\u2019t always easy. Building a machine learning training and deployment pipeline is a fractured experience from the get-go.<\/p>\n

Below, we\u2019ll go through Lightning\u2019s unified platform for training and deploying machine learning models in production.<\/p>\n

 <\/p>\n

What is Lightning?<\/h2>\n

Lightning (by the same people who built PyTorch Lightning<\/a>) is a platform that augments the capabilities of PyTorch Lightning beyond training, into serving, deploying, monitoring, and data engineering. The user-centric Lightning Trainer already made distributed training simple and gave users control over their hardware selection. Lightning was built with these same principles in mind, giving researchers and practitioners full flexibility over their work while also abstracting away complex cloud infrastructure.<\/p>\n

Lightning is designed around the Lightning App framework<\/a>, an open-source library that enables you to build and run distributed Pythonic applications. With just a few extra flags in your CLI, you can run that Python code on your local system, the Lightning Cloud (backed by AWS), or even your own private cluster\"tooltipGot AWS credits kicking around? \ud83d\ude09<\/span><\/span><\/span>\u00a0with minimal friction. Lightning brings the same robust simplicity for which PyTorch Lightning is loved to our fully-managed cloud service.<\/p>\n

Is this the fastest way to build an ML pipeline?<\/strong><\/em><\/p><\/blockquote>\n

We think so! For instance, you control infrastructure provisioning with the Lightning CloudCompute<\/a> Python API. Anyone from your machine learning team can then customize and manage the infrastructure without needing expertise in cloud computing. This again removes the friction and dependency delay<\/a> on cloud engineers, which is a time-consuming process for most organizations.<\/p>\n

 <\/p>\n

Lightning APIs<\/h2>\n

In the following section, we\u2019ll go through the core Lightning APIs\"tooltipLightningWork and LightningFlow<\/span><\/span><\/span> as well as some examples of them in use.<\/p>\n

 <\/p>\n

LightningWork<\/h3>\n

This is the building block<\/a> for long-running jobs like model training, serving, and ETL.<\/p>\n

LightningWork gives users the flexibility to customize machine configurations like RAM, CPU\/GPU, and disk allocation.<\/p>\n

Here\u2019s an example of how you can create an Nvidia T4 GPU machine with a 50 GB disk size and add PyTorch<\/code> and FastAPI<\/code> as dependency requirements:<\/p>\n