What is PyTorch? PyTorch 2019-08-19 | PyTorch is an open source machine learning framework that accelerates the path from research prototyping to production deployment. This video explains the fundamental concepts behind deep learning, and how tools like PyTorch enable developers to build and deploy AI.Visit pytorch.org for more information
How to Train Neural Networks with Multiple Losses Simultaneously using TorchJD PyTorch 2026-09-21 | Join Valérian Rey from SimplexLab at PyTorch Conference North America 2026 as he breaks down multi-objective optimization. He will share practical techniques for training models with multiple loss functions simultaneously and demonstrate how to easily implement these workflows using the TorchJD library. Whether you are optimizing multi-task architectures or managing complex loss tradeoffs, this session will provide concrete steps to streamline your pipeline. Register for PyTorch Conference North America: https://hubs.la/Q04v4SL60
Scaling ExecuTorch & Disaggregating LLM Serving PyTorch 2026-09-20 | At PyTorch Conference North America in San Jose, Arkadip Maitra from the PyTorch engineering team at Red Hat will present two technical poster sessions covering key topics in inference performance and testing infrastructure. Register for PyTorch Conference America today: https://hubs.la/Q04v4SL60
A Tiered Cross-Repository CI Relay for Out-of-Tree Backends PyTorch 2026-09-19 | At PyTorch Conference North America 2026 in San Jose, Red Hat engineers Jewel Muraledharan and Subin George Malana will present on Cross-Repository CI Relay (CRCR) to explore how downstream repositories can seamlessly integrate into upstream PyTorch CI. Their session will detail how CRCR enables a tiered onboarding model alongside real-time visibility for downstream repositories in HUD. Join us in San Jose on October 20 and 21 to learn how CRCR streamlines testing workflows across the ecosystem.Register today: https://hubs.la/Q04v4SL60
PyTorch Walkthrough of Muon, Dion, and Orthogonalized Optimizer Variants PyTorch 2026-09-18 | Muon has attracted a lot of attention for fast convergence, but getting those optimizers to work in a real training stack is a different challenge. At PyTorch Conference North America, Jen Wei will walk through Muon, Dion, and Dion3 and look at what happens when you add the messy details: sharding, communication, schedulers, and everything else that can go wrong. If you are interested in optimizers and training dynamics, see you at PyTorchCon NA. Get your ticket: https://hubs.la/Q04v4SL60
PyTorch 2.14 Release Live Q&A PyTorch 2026-09-17 | PyTorch 2.14 introduces updates across compilation, distributed training, performance, dynamic shapes, Apple Silicon, and accelerator platforms. Highlights include NVGEMM, a CuTeDSL-generated GEMM backend for Inductor; the new nccl2 backend for PyTorch Distributed; fault-tolerant process-group reconfiguration in c10d; native linear algebra on Apple Silicon; and declarative dynamic shapes with @dynamic_spec.Bring your questions about the release to our live Q&A. Andrey Talman (Meta), Nikkita Shulga (Thinking Machines Lab), Joe Spisak (Reflection AI), and Chris Gottbrath (Gottbrath Tech, moderator) will share an overview of PyTorch 2.14 and answer community questions about PyTorch and the new capabilities in the release.Topics will include:- NVGEMM and CuTeDSL-generated CUTLASS kernels in Inductor- The new nccl2 backend for PyTorch Distributed- Fault-tolerant collectives and process-group reconfiguration in c10d- Native linear algebra and additional Metal kernel improvements on Apple Silicon- torch.switch and CUDA graph capture for torch.while_loop- Declarative dynamic shapes with @dynamic_spec- Experimental torch.compile support for complex-valued tensors- Expanded ROCm, Intel XPU, and NVIDIA platform supportPyTorch 2.14 includes 2,995 commits from 487 contributors since PyTorch 2.13. The release includes work across compilation, distributed communication, device support, and accelerator platforms. PyTorch Conference North America 2026 takes place October 20–21 in San Jose, with sessions spanning compiler and runtime work, distributed communication, device portability, release engineering, CI, observability, accelerator integration, contributor infrastructure, and more. Explore PyTorch Conference North America 2026.
Retrieval as Inference: Building a Production-Scale Torch-Native Retrieval Engine with Dhritiman Das PyTorch 2026-09-17 | Dhritiman Das, Engineering Manager - AI Infrastructure at LinkedIn, will present a torch-native retrieval engine that uses PyTorch as the primary runtime for retrieval, filtering, and ranking at PyTorch Conference North America 2026.The system combines a GPU-resident torch tensor index, tensor-based retrieval operations, custom CUDA kernels for attribute filtering, and TorchScript model execution within a unified serving architecture orchestrated through a Rust and tch-rs backend.Dhritiman will also share how the approach scales to critical use cases like feed and search at LinkedIn.Join us in San Jose on October 20-21: https://hubs.la/Q04v4SL60View the poster sessions: events.linuxfoundation.org/pytorch-conference-north-america/program/schedule-posters#PyTorchCon
Federated Learning for Medical Imaging at PyTorch Conference 2026 PyTorch 2026-09-16 | Sudhan Shushawasa, a graduate student researcher in Electrical and Computer Engineering at UC Davis, discusses how federated learning can enable collaborative medical imaging research without centralizing data, using glaucoma imaging as a case study.The session covers practical lessons from training across diverse institutional datasets and building privacy-aware AI systems for medical imaging.Register for PyTorch Conference North America 2026, October 20–21 in San Jose: https://hubs.la/Q04w5M9L0
Interested in Al & compute? Join the brightest minds in the Al industry at PyTorch Conference PyTorch 2026-09-15 | New chips are shipping faster than ever before, but the ecosystem is being held back by having to rewrite and then re-debug the same code over and over again.Chris Lattner, CEO and Co-founder, Modular and Executive Vice President of Advanced AI Software and Platforms, Qualcomm, will deliver a keynote at PyTorch Conference North America about an open software platform for heterogeneous compute powered by Mojo and MAX.PyTorch has always been the place where the best models come together, and now there's a way to get those models onto all kinds of hardware. If you're interested in Al and compute, join us at the PyTorch Conference in San Jose, CA. Register now: https://hubs.la/Q04v4SL60
PyTorchCon: Where Open Innovation in AI Happens PyTorch 2026-09-14 | In this video, Ankit Patel from NVIDIA shares his perspective on the open innovation that takes place across the PyTorch ecosystem during PyTorch Conferences. Register today to join the open source AI community at PyTorch Conference North America in San Jose, October 20–21: https://hubs.la/Q04tBgv_0
Empowering developers through collaboration, choice and open contributions PyTorch 2026-09-14 | The open source AI community thrives on collaboration, choice, and open contribution. As Red Hat's Joseph Groenenboom highlights, empowering developers to freely contribute and build options is essential for growing the entire AI ecosystem.PyTorch Conference North America 2026 is where that community comes together in person. Join us to collaborate with open source pioneers, expand your deployment capabilities, and directly contribute to shaping the future of machine learning.Register for PyTorch Conference North America today: https://hubs.la/Q04tBgv_0
PyTorch can extend language modeling beyond classification to practical time-to-event prediction PyTorch 2026-09-14 | Witold Czubala of UBS will share a PyTorch-based transformer model that predicts six-month attrition from longitudinal advisor–client email communications for wealth management during PyTorch Conference North America in San Jose, CA.Join us October 20-21: https://hubs.la/Q04v4SL60
Optimizing LLM Inference: Disaggregated Serving, PD Protocol, & KV Pinning PyTorch 2026-09-11 | At PyTorch Conference North America 2026, Nicolò Lucchesi, Research Engineer at Mistral AI and vLLM maintainer, will present joint work with Amazon Web Services (AWS) and Red Hat on how disaggregated serving in vLLM has evolved to support the latest generation of hybrid models.Join us in San Jose on October 20-21: https://hubs.la/Q04v4SL60
AI Agents for LLM Inference Runtimes on Edge Hardware PyTorch 2026-09-11 | At PyTorch Conference North America, Thomas Cottenier from Arm will present how AI agents can synthesize customized, target-specific runtimes using PyTorch components like torch.export, ExecuTorch, and torchao. This approach removes redundant prefill computation and creates streamlined inference pipelines that deliver high performance directly on edge hardware. Join us San Jose this October 20-21 to learn more: https://hubs.la/Q04v4SL60
Optimizing Multi-Stage AI Pipelines: Prefix Caching for Faster Autoregressive Inference PyTorch 2026-09-10 | At PyTorch Conference North America, Ricardo Noriega de Soto, Tech Lead for the vLLM Omni team and Alexander Brooks, Principal Machine Learning Engineer at Red Hat, will demonstrate how extending vLLM’s prefix caching mechanism to multistage pipelines boosts inference speeds while reducing GPU memory overhead. Join us in San Jose on October 20th to learn practical strategies for optimizing complex AI workloads: https://hubs.la/Q04v4SL60
Why PyTorch Foundation Ambassador Arun Bhandari thinks you should attend PyTorch Conference NA 2026 PyTorch 2026-09-08 | PyTorch Conference North America 2026 offers an excellent opportunity to connect with researchers, developers, and AI engineers from around the world.In this video, Arun Bhandari, PyTorch Foundation Ambassador and leader of PyTorch Community Nepal, shares why he is excited to connect with the community and explore the latest developments in PyTorch.Register here to join us in San Jose, CA this October 20-21: https://hubs.la/Q04v88dJ0
Building high-performance recommendation inference systems using PyTorch PyTorch 2026-09-07 | At PyTorch Conference 2026, Lu Fang, Research Scientist, and Ilina Mitra, Engineering Manager at Meta, will present a deep dive on building high-performance recommendation inference systems.They will detail end-to-end production workflows, covering graph capture, model splitting, and serving across large-scale platforms like Facebook and Instagram. Attendees will gain technical insights into advanced optimizations for high TPS and low latency, as well as strategies for multi-accelerator support to handle diverse machine learning workloads at scale.Register now and connect with the open source AI community in San Jose, CA, October 20-21: https://hubs.la/Q04w5M9L0
Learn how PyreFly enables end-to-end static shape coverage for PyTorch models. PyTorch 2026-09-07 | Ever wondered why type systems don't automatically track tensor shapes across your neural network modules? In his session at PyTorch Conference North America 2026, Avik Chaudhuri from Meta will introduce Pyrefly, a static type checker for Python which enables end-to-end static shape coverage for PyTorch models. Join us in San Jose, CA, October 20-21 to explore the future of open source AI and the impact of PyTorch Foundation projects like PyTorch, vLLM, DeepSpeed, Ray, Helion, and Safetensors. Register now: https://hubs.la/Q04w5M9L0
Learn how to grow a local PyTorch community at PyTorch Conference North America 2026. PyTorch 2026-09-07 | PyTorch Conferences bring developers together from all around the world. At PyTorch Conference North America in October 2026, PyTorch Foundation Ambassadors Jiho Kim and Junghwan Park will share how they grew the PyTorch Korea community through translation, mentorship, and regional events. Register for PyTorchCon NA and connect with the global PyTorch community: https://hubs.la/Q04v88dJ0
Richard Liaw (Ray) PyTorch Conference North America 2025 PyTorch 2026-09-03 | Richard Liaw of Ray discusses the project’s focus on building useful, critical AI infrastructure and the importance of Ray within the broader AI stack. He also speaks about maintaining Ray as an open, well-governed project that continues to evolve with its community.PyTorch Conference North America returns October 20–21, 2026, in San Jose.Register: https://hubs.la/Q04w5M9L0
Primus Tuning Hybrid Projection and Agentic Search for Distributed Training PyTorch 2026-09-01 | How do you train a 671-billion-parameter model across a thousand GPUs, and do it efficiently? It comes down to a lot of choices. How do you split the model across the GPUs? What batch size? How do you schedule the work? Normally, finding the best setup means running option after option, burning weeks and a huge amount of compute. Join Anshu Raina and Peyman Razaghi of AMD at PyTorchCon NA where they show how the Primus Tuning Agent predicts how fast each option will run, ahead of time, from quick hardware benchmarks, so you find the best configuration without running them all, saving thousands of GPU-hours of trial-and-error. Register for PyTorchCon North America today: https://hubs.la/Q04v4SL60
PyTorch Conferences are where the open source AI community gathers to decide whats next PyTorch 2026-09-01 | Attendees of PyTorch Conference Europe earlier this year share what it's like to attend a PyTorchCon full of great interactions, sharing interesting ideas, thoughtful discussions, and a lot of learning.PyTorch Conference North America (October 20-21, 2026 in San Jose, CA) will be here before you know it. Don't miss out. Register now: https://hubs.la/Q04w5M9L0
Accelerating Open Ingelligence PyTorch 2026-09-01 | Join us in San Jose, CA, October 20-21 for PyTorch Conference North America 2026. This two-day event hosted by the PyTorch Foundation gathers top-tier AI pioneers, researchers, and developers to explore the future of open source AI and the impact of PyTorch Foundation projects like PyTorch, vLLM, DeepSpeed, Ray, Helion, and Safetensors.PyTorch Conference features in-depth technical talks, hands-on workshops, and candid conversations spanning the full AI stack, from bare metal infrastructure to applications and agent-based systems. The program features keynote sessions from leading voices in AI and practical deep dives on training, inference, applications across the ecosystem, kernel engineering, responsible AI, and more.PyTorch Conference is where the open source AI community connects, learns, and shapes what comes next. Register now: https://hubs.la/Q04w5M9L0
AI is an exciting field, but it is full of challenges. PyTorch 2026-09-01 | Hear from Niles Burbank (AMD) as he explains why he thinks the PyTorch Conference is unique in bringing the open source AI community together from users, to practitioners, to developers from across the AI spectrum in a way that no other event does.Join us in San Jose, CA, October 20-21 for PyTorch Conference North America 2026 to explore the future of open source AI and the impact of PyTorch Foundation projects like PyTorch, vLLM, DeepSpeed, Ray, Helion, and Safetensors. Register now: https://hubs.la/Q04w5M9L0
Meet the people behind the open source tools you use every day at PyTorchCon North America PyTorch 2026-09-01 | PyTorch Conferences present at great opportunity to meet the people behind the open source tools we use every day. PyTorch Ambassador, Abdulsalam Bande is looking forward to connecting with the wider PyTorch Community at PyTorch Conference North America this October 2026.Register for PyTorchCon NA: https://hubs.la/Q04v88dJ0
Efficient Pretraining of LLMs in NVFP4 at PyTorchCon PyTorch 2026-09-01 | Anjulie Agrusa, Ryan Spring, and Bruce Zitelli of NVIDIA will show how the latest NVFP4 pretraining recipes can accelerate large-scale LLM training while closing the quality gap to BF16 at PyTorch Conference North America 2026.“Efficient Pretraining of LLMs in NVFP4,” will cover the recipe and PyTorch tooling that make this possible, including recipe design, kernel choices, and the API surface needed to bring NVFP4 training into native PyTorch workflows.The recipes are being upstreamed into the PyTorch ecosystem through TorchAO and TorchTitan, with dense linear NVFP4 training available in TorchAO today.Register for PyTorchCon North America: https://hubs.la/Q04v4SL60
Tackle the systems challenges of Agentic RL at PyTorchCon North America PyTorch 2026-08-31 | Agentic RL introduces new systems challenges beyond traditional single-turn RL. At PyTorch Conference North America, Yichuan Wang and Shuhua Yu will discuss how to build an end-to-end agentic RL training loop in PyTorch, covering rollout infrastructure, trainer–serving interaction, environment abstractions, sandbox execution, scheduling strategies, and the tradeoffs between on-policy and off-policy training. They will examine key design choices, practical engineering considerations, and emerging techniques and recipes for scaling multi-turn agent training, drawing lessons from recent open-source and industry systems.Register for PyTorchCon North America today: https://hubs.la/Q04v4SL60
PyTorch Conference 2025: Christian Jacobi (IBM) PyTorch 2026-08-28 | Christian Jacobi (IBM) discusses how enterprises are thinking about deploying AI across core IT and business processes, including the challenge of doing so at scale with the qualities of service expected from classical enterprise computing.#PyTorchCon North America returns October 20–21, 2026, in San Jose.Register by September 4 to save on your conference pass: https://hubs.la/Q04vsCf70
PyTorch Conference 2025: Simon Mo (vLLM, Inferact) PyTorch 2026-08-28 | Simon Mo (vLLM, Inferact) discusses vLLM’s focus on ease of use and efficiency, the cost and complexity of serving large language models, and the role of open source systems in improving access and performance.#PyTorchCon North America returns October 20–21, 2026, in San Jose.Register by September 4 to save on your conference pass: https://hubs.la/Q04vsCf70
PyTorch Conference 2025 Jana van Greunen (Meta) PyTorch 2026-08-26 | Jana van Greunen (Meta) reflects on PyTorch’s role in AI development in this clip from #PyTorchCon NA 2025.#PyTorchCon North America returns October 20–21, 2026, in San Jose.Register by September 4 to save on your conference pass: https://hubs.la/Q04vsCf70
Learn how to train more with less at PyTorch Conference North America 2026 this October PyTorch 2026-08-25 | Hyungyo Kim, Ph.D. Candidate in Electrical and Computer Engineering, University of Illinois at Urbana-Champaign invites you to Poster Presentation on Train More with Less: Fewer GPUs, Less Data Movement, Better Cluster UtilizationRegister for PyTorchCon North America today: https://hubs.la/Q04v4SL60
PyTorch Conference brings together so many of the smartest people in the world PyTorch 2026-08-24 | Mark Collier, Executive Director of the PyTorch Foundation invites you to join us at PyTorch Conference North America in San Jose, October 20th - 21st, 2026.Register today: https://hubs.la/Q04v4SL60
PyTorch Conference 2025 Tristan Rice (Meta) PyTorch 2026-08-19 | Meta's Tristan Rice shares the importance of IRL community connection at PyTorch Conference NA. Register today: events.linuxfoundation.org/pytorch-conference-north-america/?
PyTorch Conference North America 2025 Sharon Zhou, PhD (AMD) PyTorch 2026-08-19 | Hear from Sharon Zhou, PhD (AMD), briefly discuss the value of PyTorch Conference North America 2025.
PyTorch 2.13 Release Live Q&A PyTorch 2026-07-23 | PyTorch 2.13 introduces updates across attention, compilation, distributed training, memory efficiency, Python support, and accelerator platforms. Highlights include FlexAttention support on Apple Silicon with up to approximately 12x speedup over SDPA on sparse patterns, the CuTeDSL "Native DSL" backend for key GPU operations, and nn.LinearCrossEntropyLoss to reduce peak GPU memory by up to 4x during large-vocabulary language model training.On Wednesday, July 22, 2026, at 11 a.m. PT, PyTorch maintainers and contributors will provide a brief overview of the PyTorch 2.13 release and answer questions from the community live.Topics will include:- FlexAttention support on Apple Silicon and deterministic backward computation on CUDA- The CuTeDSL "Native DSL" backend for Inductor- nn.LinearCrossEntropyLoss for reducing peak GPU memory- torchcomms for large-cluster training- FSDP2 communication overlap improvements- Torch wheel support for Python 3.15 on Linux, including free-threaded 3.15t builds- Expanded ROCm, Arm, and Intel XPU platform supportThe live Q&A will feature expert panelists Alban Desmaison, Andrey Talman, and Piotr Bialecki, with Chris Gottbrath moderating.PyTorch 2.13 includes 3,328 commits from 526 contributors since PyTorch 2.12.Register for the live Q&A today!
What PyTorch Conference Europe 2026 Was Really Like – Official PyTorchCon EU Highlights | Paris PyTorch 2026-05-29 | What was PyTorch Conference Europe 2026, held on April 7–8 in Paris, France, really like? Hear from attendees about their experiences.PyTorchCon EU featured inspiring keynotes, cutting-edge technical sessions, hands-on workshops, and vibrant discussions shaping the future of AI and deep learning.From major announcements and deep technical insights to powerful community moments and the unforgettable energy of Paris, here are the highlights from the first PyTorch Conference in Europe.Whether you joined us in person or are catching up now, we hope these moments inspire your next breakthrough.► View all the session recordings here youtube.com/playlist?list=PL_lsbAsL_o2DZbCiISSjbucDOW0QOxwBF #PyTorchConSee you at PyTorch Conference China and PyTorch Conference North America later this year!
PyTorch 2.12 Release Live Q&A PyTorch 2026-05-21 | PyTorch 2.12 includes major updates across compilation, distributed systems, export, graph capture, and accelerator support. Highlights include a new device-agnostic torch.accelerator.Graph API, up to 100x faster batched eigenvalue decomposition on CUDA, support for microscaling quantization formats in torch.export.save, and expanded CUDA, ROCm, XPU, MPS, and Arm platform support.Join us on Wednesday, May 20 at 10:00 AM PT for a live Q&A with panelists Andrey Talman, Alban Desmaison, and Joe Spisak, moderated by Chris Gottbrath. The panel will provide a brief overview of the release and answer your questions live. Register today!Topics include:-Device-Agnostic Accelerator Graph Capture-ProcessGroup Support in Custom Ops-torch.export.save Support for Microscaling Quantization Formats-Fused Adagrad Optimizer Support-FlightRecorder Updates-Multi-GPU and Multi-Node Profiling Improvements-Updated Backend Selection for torch.linalg.eigh on CUDA-Expanded CUDA, ROCm, XPU, MPS, and Arm Platform SupportRegister today.Panelists:Andrey Talman is a Software Engineer at Meta, primarily focused on open source releases for PyTorch and its ecosystem libraries. He works on release management, continuous integration, and process improvements, ensuring high-quality and timely delivery of PyTorch and related projects.Alban Desmaison is a Research Engineer at Meta and the Lead Core Maintainer of PyTorch.Joe Spisak is Vice President of Product and Head of Open Source at Reflection AI. He is a PyTorch core maintainer, serves on the PyTorch Foundation Governing Board, and previously worked at Meta.Moderator:Chris Gottbrath is a Group Technical Program Manager supporting PyTorch at Meta and Chair of the PyTorch Foundation Marketing Committee.
2026 PyTorch Docathon PyTorch 2026-05-06 | The 2026 PyTorch Docathon will take place May 5 to May 19, a hackathon-style community event focused on improving PyTorch documentation.Whether you're new to open source or an experienced contributor, the Docathon offers opportunities to improve tutorials, guides, examples, and website content. Many issues are beginner-friendly, with tasks labeled by skill level and support available through the PyTorch Discord.Contributors can make an immediate impact while learning more about PyTorch modules, tutorials, and workflows.ScheduleMay 5Kickoff and Q&A, 10:00 AM PTMay 6 to May 15Submissions and feedbackMay 16 to May 18Final reviewsMay 20Winner announcementsRSVP to receive updates, participation details, and event instructions.
Why WideEP Inference Needs Data-Parallel-Aware Scheduling - Maroon Ayoub & Tyler Michael Smith PyTorch 2026-04-20 | Why WideEP Inference Needs Data-Parallel-Aware Scheduling - Maroon Ayoub, IBM; Tyler Michael Smith, Red HatWideEP—wide expert parallelism fails not because experts are expensive, but because routing ignores where state already lives. In PyTorch LLM serving with vLLM, WideEP fans tokens across many experts while KV caches accumulate unevenly across data-parallel replicas. When routing is unaware of KV placement and per-replica load, requests land on replicas that cannot reuse cache or make progress efficiently and latency spikes as expert fan-out grows.The fix is not reshaping expert parallelism, but making routing data-parallel aware using signals vLLM already exposes. In this talk, we show how llm-d extends its router to leverage KV-cache locality and load awareness when routing WideEP flows. Rather than treating replicas as interchangeable, the router prefers replicas with warm KV state and available capacity, aligning routing decisions with vLLM’s execution reality and reducing cache fragmentation.This session walks through how KV-aware, data-parallel routing changes WideEP inference in practice: which signals matter, how routing behavior evolves, and where the gains come from. Attendees leave with a clear mental model for when KV- and load-aware routing unlocks higher throughput.
Lightning Talk: How DeepInverse Is Solving Imaging in Science and H... Andrew Wang & Minh Hai Nguyen PyTorch 2026-04-20 | Lightning Talk: How DeepInverse Is Solving Imaging in Science and Healthcare With PyTorch - Andrew Wang, DeepInverse; Minh Hai Nguyen, Université de ToulouseDeep learning has revolutionised imaging, a foundation of science and healthcare. DeepInverse is the PyTorch library for solving imaging problems, unifying deep learning methods (e.g. diffusion models), physics (medical, optics) and modern tooling. In this talk, we’ll show how the PyTorch community can get involved in this exciting yet accessible application of open-source AI.AI methods in imaging must model the imaging physics, leading to interesting engineering problems e.g. efficient differentiable ops, physics-informed losses. We’ll show notebooks on real use-cases: accelerating brain MRI, reducing radiation in CT scans, imaging black holes.PyTorch enthusiasts at any level/background can contribute - from training infra for scientific data to high-level generative modelling frameworks - their AI engineering skills can directly impact imaging across multiple fields.DeepInverse is supported by a growing international user community and proudly rooted in Paris. We’ve joined the PyTorch Ecosystem and received the Prix Science Ouverte in 2024. We’re excited to join the PyTorch Conf to celebrate the vibrant French developer community!
Write Once, Run Everywhere with Pytorch Transformers - Pedro Cuenca, Hugging Face PyTorch 2026-04-20 | Write Once, Run Everywhere with Pytorch Transformers - Pedro Cuenca, Hugging FaceThe Hugging Face transformers library is built on pure PyTorch and can be succinctly described as a model-definition framework. It provides an unified, familiar, clear and concise interface to multiple machine learning architectures across modalities.Serving and inference optimizations are not its focus.However, transformers model definitions become the de-facto reference implementations multiple other projects use. This includes training libraries, fast deployment engines such as vLLM and SGLang, and on-device libraries like MLX and llama.cpp.This session describes the path towards increasingly simpler downstream integration of transformers models into inference and deployment libraries, and how transformers and PyTorch core features enable the ecosystem to enjoy newly-released models as soon as they are released.We'll go through the journey towards easier modeling, which implies easier downstream porting and adaptation. The end-game is pure interoperability, where no code changes are required! This is now possible with vLLM and SGLang, and we'll show how. We'll end up discussing our ideas on upcoming interop features with MLX and llama.cpp.
The Science and Practice of Open and Scalable LLM Evaluations - Grzegorz Chlebus, NVIDIA PyTorch 2026-04-20 | The Science and Practice of Open and Scalable LLM Evaluations - Grzegorz Chlebus, NVIDIARapid advances in AI have expanded the range of capabilities required for successful real-world deployment. Understanding where we are in this multi-dimensional frontier is essential for accelerating innovation through effective quality assurance. Rigorous evaluation is increasingly difficult to scale as development requires testing many checkpoints across numerous benchmarks. Model comparison is further complicated by limited transparency of reported results. This talk explores challenges, best practices, and open-source tools that elevate evaluation to a core component of LLM development, delivering continuous signals across the model lifecycle.We discuss principles for standardizing evaluation methods and improving consistency through practical patterns and anti-patterns, and examples of integrating the science of evaluation directly into model development. Using Nemo-Evaluator, an open-source scalable evaluation tool, we demonstrate modular architectures that enable transparent, reproducible measurement. Finally, we show how Nemo-Evaluator supports reproducible evaluation for the Nemotron model family, helping enable one of the most open development processes in modern AI.
Lightning Talk: Deep Learning in the Wild: Embedded PyTorch for... Taraqur Rahman & Owen ODonnell PyTorch 2026-04-20 | Lightning Talk: Deep Learning in the Wild: Embedded PyTorch for Real-World Conservation Bioacoustics - Taraqur Rahman & Owen O'Donnell, OWL IntegrationsPassive acoustic monitoring is a powerful tool for wildlife conservation, but deploying deep learning models in remote rainforest environments introduces strict constraints on power, memory, and compute. In this talk, we present an end-to-end PyTorch-based pipeline for detecting and analyzing the endangered three-wattled bellbird using embedded deep learning systems.We cover the full lifecycle from audio preprocessing and model training in PyTorch to optimization and deployment on resource-constrained embedded devices. Topics include model architectures for sparse bioacoustic event detection, handling extreme class imbalance, model compression and quantization, and practical trade-offs between accuracy, latency, and power consumption.The session emphasizes real-world lessons learned deploying machine learning at the edge, where unreliable connectivity, noisy signals, and limited hardware define success more than benchmark metrics. Attendees will gain practical patterns for building and deploying PyTorch models for embedded and edge AI applications with real environmental impact.
The Token Slice: Implementing Preemptive Scheduling Via Chunked Decod... Maroon Ayoub & Kellen Swain PyTorch 2026-04-20 | The Token Slice: Implementing Preemptive Scheduling Via Chunked Decoding - Maroon Ayoub, IBM & Kellen Swain, GoogleProduction LLM serving faces a critical trade-off: while continuous batching maximizes throughput, it often sacrifices SLAs due to Head-of-Line (HoL) blocking. When long-context requests hijack the engine, tail latencies spike. Without fine-grained preemption, guaranteeing priority or fairness remains nearly impossible.We propose a solution: Chunked Decoding. By treating a fixed number of tokens as a "time slice," we bring 50 years of OS scheduling wisdom to inference. This technique decouples generation from completion, enabling a preemptive multitasking environment for LLMs.In this talk, we present a sidecar implementation for PyTorch-based servers (like vLLM) that orchestrates decoding in manageable chunks. This allows the system to pause, hold, or swap requests mid-stream without discarding the KV cache. We will share early evaluation results, discussing how varying chunk sizes impact priority handling and tail latency. Attendees will learn how a sidecar approach enables sophisticated scheduling while keeping the core engine lean—offering a blueprint for integrating preemptive scheduling into the next generation of model servers.
Enabling State-of-the-art Asynchronous Execution in Torch.compile With CUDA Streams - Michael Lazos PyTorch 2026-04-20 | Enabling State-of-the-art Asynchronous Execution in Torch.compile With CUDA Streams - Michael Lazos, MetaCUDA streams are a widely-used method for parallelizing GPU computation on NVIDIA GPUs. They have long been requested by our users and enable multiple key capabilities - overlapping communication and compute kernels, training on multiple batches in parallel and parallelizing kernels, all of which are needed for achieving SOTA training performance. Another key capability is activation offloading - this can be applied to any model to prevent OOMs by asynchronously storing activations in cpu memory until they are needed by the model.Before this work, torch.compile previously would graph break on CUDA stream contexts, which can be costly for models that utilize streams. Although workarounds exist (e.g. wrapping stream manipulation into custom ops), these solutions add complexity and create friction in the user experience. By enabling seamless CUDA stream support in PT2, we allow our users to leverage the familiar eager APIs for stream assignment and synchronization directly within torch.compile. This not only simplifies the workflow but also ensures that models using custom streaming patterns can run efficiently out-of-the-box without manual intervention or code restructuring.
Lightning Talk: ExecuTorch on Microcontrollers: Deploying PyTorch To... RJ Ascani & Matthias Cremon PyTorch 2026-04-20 | Lightning Talk: ExecuTorch on Microcontrollers: Deploying PyTorch To the Smallest Edge - RJ Ascani & Matthias Cremon, MetaExecuTorch extends PyTorch's reach to the most resource-constrained devices: microcontrollers, DSPs, and specialized neural processing units powering always-on sensors, wearables, and embedded systems. In this talk, we'll share the current state and roadmap for running ExecuTorch on platforms where every kilobyte of memory and milliwatt of power matters.What you'll learn:- How ExecuTorch's design enables deployment from ultra-low-power MCUs to DSP and NPU accelerators, all from a single PyTorch workflow- The state of backend support for Cadence DSPs, ARM Ethos-U and Cortex-M- Practical considerations for deploying models with sub-megabyte footprints and milliwatt power budgets- Case studies spanning always-on audio, embedded vision, and TinyML applications
Lightning Talk: TerraKit: Standardising AI-Ready Geospatial Data... Rosie Lickorish & Romeo Kienzler PyTorch 2026-04-20 | Lightning Talk: TerraKit: Standardising AI-Ready Geospatial Data Preparation for the TorchGeo Ecosystem - Rosie Lickorish & Romeo Kienzler, IBMWith the advent of geospatial foundation models, unexplored use cases are emerging that require well-curated datasets. Currently, no standardised approach exists for creating such AI-ready geospatial datasets. In this session, we introduce TerraKit: a comprehensive open-source Python library for retrieving, and processing geospatial data, that seamlessly integrates with upstream geospatial model training libraries such as TorchGeo or TerraTorch.From raster/vector annotations, TerraKit will match, download, process, align and split the requested data source (e.g., EarthData, CDSE, Planetary Computer) based on user specifications provided by a simple configuration file. TerraKit also supports spatial train/val splits and exports datasets in standard formats such as TACO datasets. TerraKit streamlines the pipeline from raw EO data to AI-ready datasets, accelerating the development of custom geospatial applications, and ensuring query and processing pipelines are reproducible. By lowering the barrier to entry, a wider community of TorchGeo and TerraTorch users are empowered to leverage foundation models for Earth observation.
Lightning Talk: Bayesian Neural Networks With Variational Inference in PyTorch - Lars Heyen PyTorch 2026-04-20 | Lightning Talk: Bayesian Neural Networks With Variational Inference in PyTorch - Lars Heyen, Karlsruhe Instute of Technology, Scientific Computing CenterUncertainty quantification is becoming more and more important as neural networks are used for increasingly critical tasks. Bayesian neural networks (BNNs) inherently provide a measure of their own uncertainty, but can be either hard to implement or inflexible if one uses common frameworks. In this session I discuss how to efficiently implement BNNs using Variational Inference within PyTorch and present torch_blue, a light-weight open source library that implements these methods with the goal of being easy to pick up, yet flexible enough for research on BNNs.
Lightning Talk: Why Your Forecasting Transformer Isn’t Working (And How To Fix It... Rosheen Naeem PyTorch 2026-04-20 | Lightning Talk: Why Your Forecasting Transformer Isn’t Working (And How To Fix It in Python) - Rosheen Naeem, Open Climate FixRenewable energy is clean — but it’s also inherently variable. Solar PV generation can change dramatically within minutes due to cloud cover and weather conditions, making accurate short-term forecasts essential for grid stability, energy trading, and smart-home optimisation.Open Climate Fix builds open and high-impact forecasting tools to accelerate the transition to a low-carbon energy system. One of these projects is Open Quartz Solar Forecast: an open-source model that uses public PV generation data, site metadata, and numerical weather prediction variables to forecast solar power for any location.In this talk, I’ll present a real case study from my Google Summer of Code project where I implemented and trained a Temporal Fusion Transformer for multi-horizon solar forecasting. I’ll cover the practical engineering challenges behind making transformer forecasting work in Python: building continuous training windows, aligning weather forecast steps with observations, separating static vs time-varying features, and stabilising training using PyTorch Forecasting and PyTorch Lightning.Attendees will leave with reusable patterns for real-world time-series forecasting pipelines.
Build PyTorch to Understand PyTorch - Vijay Janapa Reddi & Andrea Mattia Garavagno PyTorch 2026-04-20 | Build PyTorch to Understand PyTorch - Vijay Janapa Reddi, Harvard University; Andrea Mattia Garavagno, University of GenoaPyTorch's success depends on more than users—it needs engineers who understand what's inside. Engineers who can debug framework issues, optimize at the systems level, contribute upstream, and build what comes next. But ML education today produces practitioners who call APIs without understanding them. They train models without knowing why Adam needs 3× the memory of SGD, or what happens when they call loss.backward().TinyTorch is a 20-module open-source curriculum that closes this gap. Students construct PyTorch's core components—tensors, autograd, optimizers, CNNs, transformers—in pure Python, building a complete framework where every operation is code they wrote. By the final module, they don't just use PyTorch; they understand how to build it.The curriculum uses progressive disclosure, systems-first profiling from Module 01, and build-to-validate milestones—recreating ML breakthroughs from Perceptron (1958) through Transformers (2017), culminating in MLPerf-style benchmarking.TinyTorch is how we grow the next generation of PyTorch contributors and the engineers who will build what comes after.Open source: mlsysbook.ai/tinytorch
Lightning Talk: Jigsaw: Domain and Tensor Parallelism for High-Resolution Inp... Deifilia Kieckhefen PyTorch 2026-04-20 | Lightning Talk: Jigsaw: Domain and Tensor Parallelism for High-Resolution Input Training - Deifilia Kieckhefen, Karlsruhe Institute of TechnologyDistributed neural network training frameworks typically optimize for specific architectures while minimizing communication overhead. Transformer layers can be efficiently parallelized, but other operations such as convolutions often remain inefficient. This creates bottlenecks for complex model architectures.Moreover, existing tensor parallelism strategies typically replicate input data across all processes, creating redundant I/O that scales poorly with input size. In applications with heavy I/O demands-weather forecasting, medical imaging, or video processing-unsharded input data creates additional data-loading bottlenecks that could benefit from parallelization.Jigsaw is a PyTorch library that shards both model weights and input data across parallel processes. It maintains a PyTorch-like interface while parallelizing activations, convolutions, linear layers, and attention through a distributed matrix multiplication backend. We demonstrate the usability of Jigsaw across a wide range of model architectures and shows performance when scaling multi-billion-parameter models sharded across up to 8 processes and compares the scalability to DDP, FSDP, and Megatron-LM approaches.
