Last week Luca Antiga ⚡️ and Thomas Viehmann took a deep dive profiling to figure out which kernels apply to their optimization task and which are less of a fit. See episode 13 in full here: bit.ly/48tDVS9

Join the Thunder Sessions today! ⚡️ We're going to make Hugging Face Llama 3.2 1B go seriously brrr 🏎️ It's going to be fun! YouTube link 👇

Excited to raise $50M from NVIDIA, Cisco and JP Morgan! In just 12 months of Lightning Studio launched we have: - 230k+ users - over 3,000 companies - 3+ million developers on PyTorch Lightning And we’re just getting started! techcrunch.com/2024/11/21/lig…

We just raised 50M from Cisco Investments, J.P. Morgan, K5 Global and NVIDIA 🚀 It’s been an amazing ride, but we’re really just getting started ⚡️ So lucky to be a part of it, thanks William Falcon ⚡️ and team 💜 prnewswire.com/news-releases/…

🎉 The wait is finally over! 🎉 🚀 Step into the Lightning AI ⚡️ Studio and explore the ultimate guide to deploying the powerful Jina AI CLIP V2 with LitServe! 🔗 Check it out now: lightning.ai/bhimrajyadav/s… #AI #ML #JinaAI #LightningAI #Embeddings #Multilingual #Multimodal

One year ago we started to train jina embedding v1, we only have 2 full time and 1 intern. Managed to build our distributed infra using Lightning AI ⚡️ Fabric, today Fabric has powered v2, v3, part of Jina-CLIP-v2 (together with OpenCLIP). Now I’m super proud and appreciate to see

Really digging what the Jina team has done over the last few months. This is what focus, grit and ability to ship fast and high quality produces. It’s awesome to learn that they are also a Lightning AI ⚡️ Fabric shop ⚡️

We're thrilled to see our COO, Priya Shivakumar, featured in Capacity Media's latest article, "Women Leading the Charge in Open Source AI." Her leadership and vision are driving innovation and inclusivity in the AI community. Read more about her journey ➡️ bit.ly/48XRNEN

Proof that a small, focused team with the right tools can achieve incredible things! Using the Lightning stack, Jina AI scaled their distributed infra and trained multiple embedding models - all with just 2 engineers and 1 intern. 2 tools you need to know about

Supercharging Training using float8 and FSDP2 ⚡ Read our latest blog to find out how we achieve up to 50% throughput speedup while achieving loss & evaluation benchmark parity in training: hubs.la/Q02Zmk3F0

Great seeing PyTorch rolling out support for lower precision types and more flexible distributed strategies. Support coming to PyTorch Lightning and Lightning AI ⚡️ Fabric really soon (1 or 2 weeks). We're keeping it super close to the native PyTorch API for maximum flexibility.

✨ Excited to share my latest contribution to the open-source world! ✨ OpenAI compatible embeddings are now supported by Lightning AI ⚡️'s LitServe! 🎉 Big thanks to Aniket and Luca Antiga ⚡️ for their incredible guidance and support throughout the process! 🙌

*A Journey into PyTorch, the Ecosystem, and Deep Learning Compilers* Great seminar by Luca Antiga ⚡️ for Sapienza Università di Roma on the compiler's ecosystem and the recent Thunder framework by Lightning AI ⚡️! I have uploaded slides, video, & notebooks publicly. 🙂 sscardapane.it/seminars/pytor…

Join us at 11am ET when Luca Antiga ⚡️ and Thomas Viehmann chat about inference vs training when it comes to deep learning compilers—first up, training (HF inference without cudagraphs is still WIP)! Watch live ➡️ bit.ly/3Z0Xflq

📝 Planning for 2025? Save the date for the PyTorch Conference 📅 October 22-23, 2025 📍 San Francisco 🌉 #PyTorchConf

Tune in at 9am ET for the newest Thunder Session to see Luca Antiga ⚡️ and Thomas Viehmann tackle memory use. Will they manage to make it actually go down? Watch live ➡️ bit.ly/4gJLeIt

PyTorch Grows as the Dominant Open Source Framework for AI / ML 🔥 Read our #PyTorch 2024 Year in Review blog: pytorch.org/blog/2024-year…

Happy 2025 🎉 Big challenges are lying ahead of us, so wishing everyone to be their best selves in this transformative year