PyTorch Conference 2022
2022-12-02
-- online --
[PyTorch Conference 2022, PyTorch Conference 2023, PyTorch Conference 2024, PyTorch Conference 2025]
who it's for
machine learning engineers, research scientists, ml platform engineers, inference engineers, distributed systems engineers, mobile engineers, library maintainers
topics
TorchDynamo partial graph capture; TorchInductor compiler backend; torch.compile integration; symbolic shapes system; dynamic shapes support; export path and primitive ops; distributed training with FSDP and DDP; pipeline and tensor parallelism; torch::deploy subinterpreter deployment; TorchMultimodal vision language components; TorchRec table-batched embeddings; TorchRL TensorDict and replay buffer
| title | author | length |
|---|---|---|
| Opening Remarks | Aparna Ramani | 09m |
PyTorch Foundation launch moves governance to a Linux Foundation umbrella with founding members AMD, NVIDIA, Google, Amazon, Microsoft and Meta, separating… | ||
| Keynote | Soumith Chintala | 29m |
Announcement of PyTorch 2.0 unveils a new compiler pipeline: graph acquisition via Dynamo, lowering to primitives and a Python-written backend TorchInductor… | ||
| PyTorch 2.0: TorchDynamo | Peng Wu | 08m |
TorchDynamo is presented as PyTorch 2.0’s partial guarded graph capture that lets models fall back to eager when needed and supports data‑dependent control… | ||
| PyTorch 2.0: TorchInductor | Jason Ansel | 07m |
TorchInductor is a PyTorch-native compiler backend written in Python that maps PyTorch models into generated Triton kernels on NVIDIA GPUs and C++/OpenMP on… | ||
| PyTorch 2.x: What’s coming up next for export-path | Michael Suo | 10m |
A workflow that produces a single, inspectable model graph using torch Dynamo bytecode capture and export, returning explicit assumptions (shapes, device… | ||
| PyTorch 2.0: Dynamic Shapes Support | Horace He | 11m |
PyTorch 2.0 adds a symbolic shapes system that lets compiled graphs accept varying tensor sizes by propagating symbolic dimensions (implemented with sympy)… | ||
| PyTorch Distributed: Towards Large Scale Training | Anjali Sridhar | 08m |
PyTorch Distributed describes concrete scaling features: data parallelism via DDP (replicate+allreduce) and FSDP (parameter sharding with all-gather), FSDP… | ||
| Introduction to TorchMultimodal | Ankita De | 12m |
TorchMultimodal is a PyTorch library released as a beta that supplies modular components and end‑to‑end models so users can reproduce and extend multimodal… | ||
| torch::deploy: Running eager PyTorch models in production | Tristan Rice | 08m |
torch::deploy beta runs eager PyTorch models inside a C++ inference service by hosting multiple independent Python subinterpreters in one process, sharing a… | ||
| TorchRec: A PyTorch domain library for large-scale recommender systems | Dennis van der Staay | 08m |
TorchRec is a PyTorch domain library that swaps nn.EmbeddingBagCollection for sharded, module-level equivalents, adds collectives and awaitable sharding… | ||
| TorchRL: The Reinforcement Learning and Control library for PyTorch | Vincent Moens | 11m |
TorchRL presents a PyTorch-first RL library that is modular, Python-native and production-aware, with a separate open-source TensorDict data carrier that… | ||
| PyTorch Mobile: Past, present and 2.x future | RAZIEL ALVAREZ GUEVARA | 09m |
PyTorch Mobile claims production-grade on-device ML with real results: released three years ago, it runs across Meta apps with 50 models doing 70 billion daily… | ||
| nothing matches | ||