PyTorch

Have questions about PyTorch 2.6? Drop your questions in the comments and get them answered live tomorrow! PyTorch Core Maintainer Nikita Shulga (Meta) will tackle them during our Q&A session tomorrow at 10 am PST. Don’t miss this chance to hear directly from a key contributor to PyTorch releases and continuous integration. #pytorch #deeplearning #opensource

We’re thrilled to announce that our fully automated Triton warp specialization optimization will be available in the upcoming release of Triton 3.2, shipping with PyTorch 2.6! 👋 Learn more: https://hubs.la/Q035w8jG0 PyTorch users can now easily leverage this powerful feature by implementing custom Triton kernels, unlocking new levels of performance and efficiency. Stay tuned for more updates! 🔥 #pytorch #triton #machinelearning #deeplearning #ai #optimization #opensource

PyTorch remains a leading choice for deep learning, particularly in academia, due to its simplicity, flexibility, and dynamic computation graph for rapid prototyping. As Jad Khalife notes in TechRadar, "PyTorch is preferred in the academic community due to its simplicity and flexibility. He also believes its dynamic graph makes it easier to prototype and experiment with models. NLP is one of the fields where PyTorch is widely used…many AI developers look at PyTorch as the preferred tool for many NLP tasks, especially those that require training complex neural network models on large amounts of textual data." 💡Read more: https://hubs.la/Q03545tn0 #pytorch #opensource #deeplearning

New PyTorch Expert Exchange: Multi-Modal Tabular Deep Learning with PyTorch Frame. Register today! Akihiro Nitta will introduce PyTorch Frame, a modular framework for multi-modal tabular deep learning. PyTorch Frame enables seamless integration with the PyTorch ecosystem, including PyTorch Geometric for graph-based message passing across relational data and Hugging Face Transformers for extracting rich text features. The talk also highlights its specialized data structures for efficiently handling sparse features, making PyTorch Frame an essential tool for modern tabular data. Akihiro Nitta is a software engineer on the ML team at Kumo.ai and a core contributor to PyTorch Frame and PyTorch Geometric, with prior experience as a maintainer of PyTorch Lightning. Register today! #pytorch #expertexchange #opensource

PyTorch 2.6 is here! This release brings key improvements, including:
 🔥 torch.compile support for Python 3.13 🔥 A new performance tuning knob: torch.compiler.set_stance
 🔥 Enhancements to AOTInductor
 🔥 FP16 support on X86 CPUs Learn more in our release blog: https://hubs.la/Q034zSMz0 #pytorchrelease #pytorch #opensource

Explore the latest blog outlining 2025 priorities for the PyTorch Technical Advisory Council (TAC), from TAC Chair Luca Antiga ⬇️ Focus areas:
 - Open, multi-cloud CI infrastructure - Enhanced accelerator support - Establish a high-quality, user-centric ecosystem - Strengthening community and industry feedback loops 🗣️ Dive into the details and learn how you can support these goals: https://hubs.la/Q034qp600 #pytorch #ai #opensource

Wondering what's new in the recent PyTorch 2.6 release? Do you have questions? Join us for a live Q&A on PyTorch 2.6 with PyTorch Core Maintainer, Nikita Shulga (Meta). Nikita is a Software Engineer at Meta where, among other things, Nikita is responsible for PyTorch releases and continuous integration. Nikita is committed to uplifting the developer community and continuously improving PyTorch and earned a Master’s degree in Applied Mathematics from the Moscow Institute of Physics and Technology (MIPT). Bring your PyTorch 2.6 questions for Nikita during this live Q&A session.

Intel + PyTorch: Powering Generative AI on Intel Arc GPUs ⬇️ Intel’s new AI Playground showcases advanced GenAI capabilities, including image generation and chatbots, powered by PyTorch. By leveraging Intel® Extension for PyTorch, Intel optimized performance and accelerated development, delivering an open source solution for AI PCs. 📖 Read the case study: https://hubs.la/Q033-2X10 #pytorch #genai #ai #opensource

From today's webinar: Anton Pidkuiko shares how he built an AI agent that ranked #1 in the Meta HackerCup 2024 (AI division). He walks through using cutting-edge reasoning LLMs, RAG techniques and cloud infrastructure to solve complex programming problems at scale. Get a glimpse of Anton's iterative approach to refining AI solutions. For the full story and deeper insights, watch the recording: https://hubs.la/Q033Zy-M0 #pytorch #ai #expertexchange

🌎 Reflecting on 2024: A year of incredible PyTorch community events! From meetups in Seattle, Shanghai, and Korea to global hackathons, webinars and PyTorch Conference 2024, we celebrated learning, collaboration and innovation in AI. ➡️ Highlights and what's ahead in 2025: https://hubs.la/Q033DPs90 #pytorch #opensource #ai