An Open Source Infrastructure for PyTorch

In this talk we’ll go over tools and techniques to deploy PyTorch in production. The PyTorch organization maintains and supports open source tools for efficient inference like pytorch/serve, job management pytorch/torchx and streaming datasets like pytorch/data. This talk will give an overview of the tools we use both internally at Meta and recommend externally for easy MLops that can scale to hundreds of machines.

Speaker

Mark Saroufim

Applied AI Engineer @Meta

Mark is an Applied AI engineer in the Business Engineering group at Meta who spends most of his time maintaining or contributing to github.com/pytorch/{serve,pytorch,torchx,data,examples}. He's passionate about building in the open and even more passionate about online communities. 

Read more
Find Mark Saroufim at:

Date

Wednesday Dec 7 / 11:20AM PST ( 50 minutes )

Track

MLOps

Topics

Machine Learning Open-Source Infrastructure Pytorch/Serve Job Management Pytorch/Torchx Pytorch/Data Languages

Share

Share Share

From the same track

Session Machine Learning

Ray: The Next Generation Compute Runtime for ML Applications

Wednesday Dec 7 / 09:00AM PST

Ray is an open source project that makes it simple to scale any compute-intensive Python workload. Industry leaders like Uber, Shopify, Spotify are building their next generation ML platforms on top of Ray.

Speaker image - Zhe Zhang
Zhe Zhang

Head of Open Source Engineering @anyscalecompute

Session Machine Learning

Fabricator: End-to-End Declarative Feature Engineering Platform

Wednesday Dec 7 / 10:10AM PST

At Doordash, the last year has seen a surge in applications of machine learning to various product verticals in our growing business. However, with this growth, our data scientists have had increasing bottlenecks in their development cycle because of our existing feature engineering process.

Speaker image - Kunal Shah
Kunal Shah

ML Platform Engineering Manager @DoorDash

Session Machine Learning

Real-Time Machine Learning: Architecture and Challenges

Wednesday Dec 7 / 12:30PM PST

Fresh data beats stale data for machine learning applications. This talk discusses the value of fresh data as well as different types of architecture and challenges of online prediction.  

Speaker image - Chip Huyen
Chip Huyen

Co-founder @Claypot AI

Session Machine Learning

Declarative Machine Learning: A Flexible, Modular and Scalable Approach for Building Production ML Models

Wednesday Dec 7 / 01:40PM PST

Building ML solutions from scratch is challenging because of a variety of reasons: the long development cycles of writing low level machine learning code and the fast pace of state-of-the-art ML methods to name a few.

Speaker image - Shreya Rajpal
Shreya Rajpal

Founding Engineer @Predibase