Accelerate Llama3 FP8 Inference with Triton Kernels 😎

We present an optimized Triton FP8 GEMM kernel TK-GEMM, which leverages SplitK parallelization.

The kernel code is publicly available, check out the blog here: hubs.la/Q02vWwFf0

Join us June 4-20 for our third annual Docathon! 💜

Similar to a hackathon, the docathon is dedicated to enhancing the quality of the PyTorch documentation with the help of our community.

Register today: hubs.la/Q02vW8GZ0

2x-4x better inference latencies on Llama 🚀

Try out the new speculative decoding approach from IBM Research, with end-to-end recipes to train your own speculators!

Read more: hubs.la/Q02vW3gS0

📝 The #PyTorch Conf #CallForProposals is accepting talks for THE cutting-edge #OpenSource #MachineLearning framework event, Sept 18-19 in San Francisco. If you're an expert on #PyTorch framework, #DeepLearning , chatbots, or #security , submit by June 7: hubs.la/Q02vVhX00.

Introducing ExecuTorch Alpha ⚡

ExecuTorch Alpha is focused on deploying large language models and large ML models to the edge, stabilizing the API surface, and improving installation processes.

Learn more in our latest blog: hubs.la/Q02vzrrW0

Don't miss THE event covering cutting-edge #OpenSource #MachineLearning framework. 🔥 #PyTorchConf comes to San Francisco September 18-19! Learn the latest in #AI & #ML innovations & make connections w/ the best in the industry. Register by July 12 & save! hubs.la/Q02vy6YD0

🗣️ Submit to speak at #PyTorchConf in San Francisco September 18-19! We're looking for talks on the cutting-edge #OpenSource #MachineLearning framework, #DeepLearning , chatbots, #security & MORE. Submit your proposal by June 7: hubs.la/Q02ttRPp0.

PyTorch 2.3 is here 😎🔥

PyTorch 2.3 offers support for user-defined Triton kernels in torch.compile, allowing for users to migrate their own Triton kernels from eager without experiencing performance regressions or graph breaks.

Details: hubs.la/Q02tYcYq0

Don't miss our in person premiere of the PyTorch Documentary! Details below.

📍 Paris, France
🗓️ June 18 17:00 PM
🎟️ Register now: hubs.la/Q02tPT5m0

Join us in San Francisco for #PyTorchConf , September 18-19, & discover the cutting-edge #MachineLearning framework revolutionizing #OpenSource ! We'll bring you world-class content covering the future of #AI & #ML innovation. 🚀 Register by April 26 & SAVE: hubs.la/Q02tm-nZ0

Happy Earth Day! 🌍

Starting soon: Webinar on predicting wildfires with PyTorch at 10am PT/1pm ET.

Join us at the link below!
hubs.la/Q02tCV9D0

🎤 Submit to speak @ THE event covering cutting-edge #OpenSource #MachineLearning framework - #PyTorch Conf comes to San Francisco September 18-19! We need experts on #PyTorch framework, #DeepLearning , chatbots, #security & MORE. Submit your talk by June 7: hubs.la/Q02t4Nng0

Don't miss #PyTorchConf , September 18-19 in San Francisco, THE conference covering cutting-edge #OpenSource #MachineLearning framework! Immerse yourself in world-class content that is propelling #AI & #ML innovation. ⚙️ Register by April 26 & save US$350: hubs.la/Q02t4rSx0.

Announcing the alpha release of torchtune!

torchtune is a PyTorch-native library for fine-tuning LLMs. It combines hackable memory-efficient fine-tuning recipes with integrations into your favorite tools.

Get started fine-tuning today!

Details: hubs.la/Q02t214F0

🥁🥁🥁 Introducing: The premiere of the PyTorch Documentary!

Join us on June 18 in Paris to see the first ever showing of PyTorch’s inception, as told by a dedicated group of heroes driving technological innovation.

Register to secure your spot: hubs.la/Q02sqSxL0

Call for proposals for PyTorch Conference 2024 are live 😎 Submit to speak until June 7 - we can't wait to hear from you!

Full details + guidelines here: hubs.ly/Q02sr2s60

Intel AI Webinar Series Part 3️⃣

Learn how to perform image analysis to predict potential forest fire likelihoods based on regions of known forest fires acquired via the MODIS dataset using PyTorch.

Register: hubs.la/Q02sgYCW0

PyTorch Conference registration is now OPEN!

Join us in San Francisco to immerse yourself in top-tier content, connect globally with developers, and stay at the forefront of the evolving tech landscape of AI.

Secure your spot: hubs.la/Q02sg_9-0

Accelerating MoE model inference with Locality-Aware Kernel Design 🔥

Check out several different work decomposition and scheduling algorithms for MoE GEMMs and how at the hardware level, why column-major scheduling produces the highest speedup.

hubs.la/Q02rRKh00

⏰ Starting soon: Webinar on Seismic Data to Subsurface Models with OpenFWI with our friends from Intel AI!

Watch it here: hubs.la/Q02p-rcC0