ptrblck

I just posted my 10,000th reply in the @PyTorch discuss forum! Thanks everyone for creating such a great community, for the guidance and mentorship I received, and @soumithchintala for starting this journey.

@PyTorch @Nvidiadev @matthew_d_white @narendasan @PKUWZP @robertnishihara @_seemethere @mr_purelycloud @alband @malfet @joespeez @syedahmedz Wrong Syed tagged! It's @tousifsays :)

🗓️ TUESDAY @ Booth #338 12PM: Wheel Next w/ @_seemethere 1PM: OpenEnv w/ Ali Sol & @mr_purelycloud 2PM: Core Maintainer Office Hours w/ @albanD @malfet @joespeez @ptrblck_de 3PM: Titan & Forge w/ @syedahmedz, Davide Italiano, Kai Wu 4PM: Ambassador Program w/ @matthew_d_white 5PM: vLLM Inference w/ @woosuk_k

Heading to #NVIDIAGTC next week? Let’s talk @PyTorch. 🚀 We’re bringing the community to San Jose. Drop by Booth #338 to meet expert developers and core maintainers in person. Scaling, inference, foundation models, and OSS contributions. Full schedule below 👇 #PyTorch

@soumithchintala @giffmana @msharmavikram Additionally, Nsight Systems now also comes with the `--pytorch` option: #pytorch-profiling" target="_blank" rel="nofollow noopener">docs.nvidia.com/nsight-systems…

@giffmana @msharmavikram @ptrblck_de torch.autograd.profiler.emit_nvtx. It emits nvtx markers for every torch op. #torch.autograd.profiler.emit_nvtx" target="_blank" rel="nofollow noopener">docs.pytorch.org/docs/stable/au…

nsys looks pretty cool actually, but information overload for a first-time user. Took me a bit to get good at Google's XProf too, so let's get started! QQ to my nsys expert followers: any specific pro-tips? Biggest bang-for-buck things/views to look at? Any good pytorch training-specific profiling walk-through video or slides or blog you'd recommend? 25% on a thing called "fused zeros" sounds like I may be doing something very wrong still, will have to find out😅 Also, hate the scrolling zoom, no WASD navigation?

Update from the PyTorch maintainers: 2.7 is out now. 🔹 Support for NVIDIA Blackwell (CUDA 12.8) 🔹 Mega Cache 🔹 torch.compile for Function Modes 🔹 FlexAttention updates 🔹 Intel GPU perf boost 🔗 Blog: hubs.la/Q03jBPSL0 📄 Release notes: hubs.la/Q03jBPlW0 #PyTorch #OpenSourceAI

@TheGoonsta08 @PyTorch @nvidia @Meta Blackwell support is already supported in our nightly binaries if you select CUDA 12.8 for some time. PyTorch 2.7.0 will support it as well.

@PyTorch @ptrblck_de @nvidia @Meta all i want to know is if blackwell support is in the pipeline...im just now getting into AI and I bought a 50 series gpu for my first build. stable diffusion is basically impossible for a layman like myself currently

Curious about what’s coming in PyTorch 2.7? Core Maintainers @ptrblck_de (@nvidia ) and Nikita Shulga (@Meta ) will take them live during a Q&A on April 28 at 12 PM PST. Hear directly from the folks behind CUDA, CI, and releases. 🔗 hubs.la/Q03hHkFP0 #PyTorch #PyTorch27 #OpenSourceAI #ML

Our banger hackathon this Sunday Over 100 B200 / GB200 to hack on Participants from every lab Prizes of Blackwell GPUs + more Speakers - Phil Tillet OpenAI, Horace He Thinking Machines, Tri Dao Together, Vijay Nvidia, and Mark GPUMode/PT 30 spots left Apply NOW Link in next tweet

popular computer vision package ultralytics (home of yolov8 and yolo11) was compromised. a crypto miner was injected into versions 8.3.41 and 8.3.42. link: github.com/ultralytics/ul…

@soumithchintala @A_Reichenbach_ @jeremyphoward @charliermarsh @chrisbbh Yes, just confirming, as well as with NCCL: pypi.org/project/nvidia…

@A_Reichenbach_ @ptrblck_de @jeremyphoward @charliermarsh @chrisbbh yes including cudnn which is now packaged on pypi pypi.org/project/nvidia…

Oh wow. This is gonna be super tricky to figure out what to do now. There isn’t any easy automated way to install a full deep learning stack from scratch after this change afaict.

@soumithchintala @jeremyphoward @charliermarsh @chrisbbh Yes, `pip install torch` will also use and download the dependencies. PyTorch binaries always shipped with all needed CUDA runtime dependencies (previously packaged into the torch wheel).

@jeremyphoward @charliermarsh @chrisbbh @ptrblck_de and team at nvidia have been bringing cudnn, cublas etc. to PyPI. pypi.org/project/nvidia…

NetworkX from NVIDIA is one THE most popular Python graph analytics library with ~15K Github starts and 80M downloads monthly. This library is for working with networks and graphs. It helps analyze connections between things - like social networks, computer networks, or any system where objects are connected to each other. And now NetworkX just got massively accelerated after its backend integration with NVIDIA's cuGraph. ✨ Up to 500x speedups on large graph workloads in NetworkX with zero code changes. And it is Zero Code Change Acceleration. 📌 cuGraph is NVIDIA's GPU-accelerated graph analytics library within the RAPIDS ecosystem. The library provides fast graph algorithms on GPUs, supporting property graphs, remote operations, and graph neural networks (GNNs). Works with GPU DataFrames (cuDF) and integrates smoothly with NetworkX-like API. -------- 📌 The traditional bottleneck of NetworkX's pure Python implementation becomes apparent when processing graphs larger than 100K nodes and 1M edges. 📌 And so now cuGraph solves this by offloading supported algorithms to the GPU. PageRank, Louvain community detection, betweenness centrality, and about 60 other algorithms get instant acceleration. 📌 This acceleration enables previously impractical use cases. Fraud detection systems can now process massive transaction networks in real-time. Recommendation engines handle millions of user-item interactions efficiently. Social network analysis scales to entire platforms worth of data on a single machine. @NVIDIAAIDev

@rasbt @lantiga @PyTorch It was great meeting you finally! Your books and lectures were my reference while digging into ML and now I even got a signed copy of your new book! Time to build an LLM from scratch!

Such a great conference, had an awesome time! @ptrblck_de was how I learned PyTorch and it was great to finally meet! Ha, and honestly, every time someone asks an LLM a question about PyTorch today and gets a coherent answer, it’s probably thanks to his contributions to the PyTorch community. I wouldn’t be surprised if >90% of the LLM training data on PyTorch comes from his work!

@lantiga @PyTorch @rasbt It was great seeing so many people and I was definitely starstruck seeing @rasbt for the first time!

What a great recognition from @PyTorch. 🎉🏆 🎊Congrats to all winners - with an extra congrats to our very own Eddie Yan from #NVIDIA!🙌 #PyTorchConf

For those at #PyTorchConf - if you missed our meetup earlier this week, stop by our lounge on the 2nd floor. We have 40+ technical experts that can answer any questions. nvda.ws/3Xz2CHQ We hope to see you there. 🙌

📣 CUDA MODE goes live IRL! Join us in San Francisco on Sept. 21 for the #Hackathon event. 🎉 Make friends, enjoy keynotes from #CUDA experts, then hack all day and night! ➡️ Apply for a spot now: nvda.ws/3AcLAHB

Excited to share a blog series I've been working on, diving deep into CUDA programming! Inspired by the #PMPP book & #CUDA_MODE!! Check out the links below...

Are YOU ready?! @ptrblck_de of @NVIDIA is taking the stage at #PyTorchConf 2024! Explore the list of keynote speakers: hubs.la/Q02NfCYY0. Register & join us September 18-19 in San Francisco! hubs.la/Q02NfyxM0

This September, we’re hosting a CUDA MODE hackathon alongside our friends at @nvidia and @PyTorch. Join us in San Francisco to spend the day hacking on new projects and hear from some of the most well-respected ML leaders at @Meta/@PyTorch, @EurekaLabsAI, @vllm_project, @togethercompute, @nvidia, and more. ⬇️

🌟We are honored that our own Piotr Bialecki (@ptrblck_de) has been designated a #PyTorch Superhero Contributor, and now a Core Maintainer, where he will continue to help build the long-term vision for PyTorch, champion #OSS, and be a key member of the governing body and community. ➡️ nvda.ws/4d9eAhG 🎉Congrats Piotr!

You can just do stuff youtube.com/clip/UgkxIHVMD…