Mark Saroufim

This Deep Dive provides an update on PyTorch development since last conference and dives into key new features. We will explore how advancements across a number of PyTorch features combine to better support the full model development lifecycle across training, fine-tuning, and deployment.

KernelBot is a competitive platform that showcases the power of community-driven innovation, hosted on GPU MODE's vibrant ecosystem of 17K developers. Our platform democratizes state-of-the-art kernel authoring by bringing together GPU programmers from diverse backgrounds to collaboratively push...

This Deep Dive provides an update on the PT2 development since last conference and dives into the key new features coming in PyTorch 2.1 This will provide high level updates on compile, distributed, inference, export and edge.

Historically for inference, users have had to rewrite their models to be jit scriptable which required model rewrites and familiarity with C++ services. This is frustrating especially when the vast majority of real world pytorch users actually deploy python in production. When torch.compile was...

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.

How can users developing machine learning (ML) workloads start small and scale big on Amazon EC2? In this chalk talk, review architectures best suited to scale ML model training and inference.