Загрузка видео...

Не удалось загрузить видео

На главную

The Together AI kernels team pushes performance to the next level. An investigation into how left more questions than answers, but VP of Kernels Dan Fu seemed proud. If you want the full picture, read on:

togethercompute's profile picture

Together AI

57,475 subscribers

36,354 просмотров • 4 месяцев назад •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

AI that responds in under 100ms doesn't happen by accident. Together AI VP of Kernels Dan Fu shares how they use NVIDIA CUDA, TensorRT-LLM, Dynamo and Together ATLAS on NVIDIA Blackwell to power ultra-low-latency inference and long context code generation for Cursor. From building megakernels that fuse an entire model's forward pass into a single launch to enabling real-time voice agents, Together AI is optimizing latency at every layer of the stack.
1:51

AI that responds in under 100ms doesn't happen by accident. Together AI VP of Kernels Dan Fu shares how they use NVIDIA CUDA, TensorRT-LLM, Dynamo and Together ATLAS on NVIDIA Blackwell to power ultra-low-latency inference and long context code generation for Cursor. From building megakernels that fuse an entire model's forward pass into a single launch to enabling real-time voice agents, Together AI is optimizing latency at every layer of the stack.

NVIDIA AI Infrastructure

20,633 просмотров • 1 месяц назад

Introducing The AI CUDA Engineer: An agentic AI system that automates the production of highly optimized CUDA kernels. The AI CUDA Engineer can produce highly optimized CUDA kernels, reaching 10-100x speedup over common machine learning operations in PyTorch. Our system is also able to produce highly optimized CUDA kernels that are much faster than existing CUDA kernels commonly used in production. We believe that fundamentally, AI systems can and should be as resource-efficient as the human brain, and that the best path to achieve this efficiency is to use AI to make AI more efficient! We are excited to publish our paper, The AI CUDA Engineer: Agentic CUDA Kernel Discovery, Optimization and Composition. We also release a dataset of over 17,000 verified CUDA kernels produced by The AI CUDA Engineer. Paper: Kernel Archive Webpage: HuggingFace Dataset: The AI CUDA Engineer utilizes evolutionary LLM-driven code optimization to autonomously improve the runtime of machine learning operations. Our system is not only able to convert PyTorch code into CUDA kernels, but through the use of evolution, it can also optimize the runtime performance of CUDA kernels, fuse multiple operations, and even discover novel solutions for writing efficient CUDA operations by learning from past innovations! We believe The AI CUDA Engineer opens a new era of AI-driven acceleration of AI and automated inference time optimization. We (Robert Lange, Aaditya Prasad 🇺🇸, sssss, Maxence Faldor, Yujin Tang, hardmaru) are excited to continue Sakana AI's mission of leveraging AI to improve AI.
1:21

Introducing The AI CUDA Engineer: An agentic AI system that automates the production of highly optimized CUDA kernels. The AI CUDA Engineer can produce highly optimized CUDA kernels, reaching 10-100x speedup over common machine learning operations in PyTorch. Our system is also able to produce highly optimized CUDA kernels that are much faster than existing CUDA kernels commonly used in production. We believe that fundamentally, AI systems can and should be as resource-efficient as the human brain, and that the best path to achieve this efficiency is to use AI to make AI more efficient! We are excited to publish our paper, The AI CUDA Engineer: Agentic CUDA Kernel Discovery, Optimization and Composition. We also release a dataset of over 17,000 verified CUDA kernels produced by The AI CUDA Engineer. Paper: Kernel Archive Webpage: HuggingFace Dataset: The AI CUDA Engineer utilizes evolutionary LLM-driven code optimization to autonomously improve the runtime of machine learning operations. Our system is not only able to convert PyTorch code into CUDA kernels, but through the use of evolution, it can also optimize the runtime performance of CUDA kernels, fuse multiple operations, and even discover novel solutions for writing efficient CUDA operations by learning from past innovations! We believe The AI CUDA Engineer opens a new era of AI-driven acceleration of AI and automated inference time optimization. We (Robert Lange, Aaditya Prasad 🇺🇸, sssss, Maxence Faldor, Yujin Tang, hardmaru) are excited to continue Sakana AI's mission of leveraging AI to improve AI.

Sakana AI

1,160,266 просмотров • 1 год назад

It's here: We just hit superhuman performance on AI kernel optimization! Real customer models & production settings. Not toy problems (what I typically see). This is the year that Claude writes its own kernels, Codex its own kernels, for every new GPU that it wants to run on -- something that takes months to port between GPU generations today. This has a massive impact to scaling intelligence. More compute means getting the next frontier model sooner.
2:48

It's here: We just hit superhuman performance on AI kernel optimization! Real customer models & production settings. Not toy problems (what I typically see). This is the year that Claude writes its own kernels, Codex its own kernels, for every new GPU that it wants to run on -- something that takes months to port between GPU generations today. This has a massive impact to scaling intelligence. More compute means getting the next frontier model sooner.

Sharon Zhou

241,935 просмотров • 5 месяцев назад

“If you want kids to learn, they need to struggle. They need to feel a level of risk. They’re asking the questions. They’re finding the answers.” Teacher Casey Fletcher on the importance of student-centered learning.
1:00

“If you want kids to learn, they need to struggle. They need to feel a level of risk. They’re asking the questions. They’re finding the answers.” Teacher Casey Fletcher on the importance of student-centered learning.

edutopia

34,821 просмотров • 3 лет назад

Your website gets visitors… But how many leave without getting answers? 👀 Most businesses don’t have a traffic problem. They have an answer problem. Visitors want instant answers: → What do you offer? → Pricing? → How to get started? That’s where AI can change the game. I explored Daysi AI — an AI website assistant that answers visitor questions, captures leads, and helps businesses turn traffic into conversations. The future of websites isn’t just more pages. It’s smarter conversations. 🚀 Would you let AI handle your website visitors?
0:37

Your website gets visitors… But how many leave without getting answers? 👀 Most businesses don’t have a traffic problem. They have an answer problem. Visitors want instant answers: → What do you offer? → Pricing? → How to get started? That’s where AI can change the game. I explored Daysi AI — an AI website assistant that answers visitor questions, captures leads, and helps businesses turn traffic into conversations. The future of websites isn’t just more pages. It’s smarter conversations. 🚀 Would you let AI handle your website visitors?

Jay Bisen

42,833 просмотров • 2 месяцев назад

underrated gems in Kimi-K3 release: > an early K3 wrote the majority of the kernels in the late development stages > it built a triton-class compiler from scratch, MiniTriton, that delivers performance on par with or better than Triton and torch.compile > then it designed a chip, by a model, for a model, in one 48-hour autonomous run the model is rewriting and optimizing every layer of the stack it runs on: kernels for its own training. A compiler for its own kernels. silicon for its own weights. we're watching software build its own hardware
0:22

underrated gems in Kimi-K3 release: > an early K3 wrote the majority of the kernels in the late development stages > it built a triton-class compiler from scratch, MiniTriton, that delivers performance on par with or better than Triton and torch.compile > then it designed a chip, by a model, for a model, in one 48-hour autonomous run the model is rewriting and optimizing every layer of the stack it runs on: kernels for its own training. A compiler for its own kernels. silicon for its own weights. we're watching software build its own hardware

Alex Ker 🔭

310,834 просмотров • 1 месяц назад

The Opps picked up the win in their ROH debut, but it ended with more questions than answers as the Premier Athletes left BEEF to fight alone!
1:47

The Opps picked up the win in their ROH debut, but it ended with more questions than answers as the Premier Athletes left BEEF to fight alone!

ROH - Ring of Honor Wrestling

11,377 просмотров • 1 месяц назад

What if your LLM inference automatically got faster the more you used it? Introducing ATLAS from the Together AI Turbo research team. Read more: Here’s Together AI Founder and Chief Scientist Tri Dao introducing ATLAS:
1:12

What if your LLM inference automatically got faster the more you used it? Introducing ATLAS from the Together AI Turbo research team. Read more: Here’s Together AI Founder and Chief Scientist Tri Dao introducing ATLAS:

Together AI

143,646 просмотров • 10 месяцев назад

Muneeb Qadir Sahib, I raised serious questions, but you seemed more interested in mimicry than meaningful answers. Quite surprising for a self-proclaimed history buff to display such remarkable ignorance about the Kashmir dispute.
0:26

Muneeb Qadir Sahib, I raised serious questions, but you seemed more interested in mimicry than meaningful answers. Quite surprising for a self-proclaimed history buff to display such remarkable ignorance about the Kashmir dispute.

Arvind Saharan

22,161 просмотров • 6 месяцев назад

A sneak peek into what the Arena crew are working on. If you are putting more time than you should into AI, this is the kind of environment you want to be in. 📍 Sentient House, Presidio
1:27

A sneak peek into what the Arena crew are working on. If you are putting more time than you should into AI, this is the kind of environment you want to be in. 📍 Sentient House, Presidio

Sentient

19,116 просмотров • 4 месяцев назад

A passenger on a Super Metro bus from Westlands to Juja reportedly refused to pay the fare. I’m left with more questions than answers!
0:53

A passenger on a Super Metro bus from Westlands to Juja reportedly refused to pay the fare. I’m left with more questions than answers!

The Kenyan Vigilante

143,758 просмотров • 11 месяцев назад

We built an AI feature on Edubaloo to help African students solve difficult questions by taking pictures of the questions with their phones. Super proud my team 🔥💪🏽
0:45

We built an AI feature on Edubaloo to help African students solve difficult questions by taking pictures of the questions with their phones. Super proud my team 🔥💪🏽

Jay Osho

11,220 просмотров • 2 лет назад

BREAKING: MICROSOFT JUST ANNOUNCED TO BAN ITS OWN ENGINEERS FROM USING AI DUE TO THE COST OF USING IT. VP OF NVIDIA SAID, “THE COST OF AI FOR MY TEAM WAS MORE THAN HUMANS” “AI CAN COST MORE THAN HUMAN WORKERS NOW”
1:33

BREAKING: MICROSOFT JUST ANNOUNCED TO BAN ITS OWN ENGINEERS FROM USING AI DUE TO THE COST OF USING IT. VP OF NVIDIA SAID, “THE COST OF AI FOR MY TEAM WAS MORE THAN HUMANS” “AI CAN COST MORE THAN HUMAN WORKERS NOW”

Vivek Sen

17,500,225 просмотров • 2 месяцев назад

This investigation is a great example of how the FBI works with our law enforcement partners at the local, state, and federal level to keep communities safe and a reminder to the public if you see something, say something. Read more:
0:42

This investigation is a great example of how the FBI works with our law enforcement partners at the local, state, and federal level to keep communities safe and a reminder to the public if you see something, say something. Read more:

FBI

34,552 просмотров • 1 месяц назад

As some of you may have seen, we’re rolling out some changes to the way Drafts work in Figma for users on Professional + Starter teams. If you haven’t seen it yet, you will by end of the week. I put together a short video to walk you through what’s changing that will hopefully answer some of your questions. You can also read more on the blog: A few common questions and answers in the thread 👇
5:28

As some of you may have seen, we’re rolling out some changes to the way Drafts work in Figma for users on Professional + Starter teams. If you haven’t seen it yet, you will by end of the week. I put together a short video to walk you through what’s changing that will hopefully answer some of your questions. You can also read more on the blog: A few common questions and answers in the thread 👇

Tom Lowry

79,576 просмотров • 2 лет назад

Why is the #chiefs and #dolphins game exclusively steaming on Peacock? What exactly went into that decision? A lot of people have questions. The NFL's VP of Broadcast Planning Mike North was on The NFL Report to provide the answers.
3:04

Why is the #chiefs and #dolphins game exclusively steaming on Peacock? What exactly went into that decision? A lot of people have questions. The NFL's VP of Broadcast Planning Mike North was on The NFL Report to provide the answers.

James Palmer

199,739 просмотров • 2 лет назад

OpenAI engineer: "At OpenAI, 85% of our engineers are running hundreds of agents with Codex. The way you do it is graph engineering." In 30-minutes lecture, he explains how a single engineer can now do the work of an entire team, how far this has already gone inside OpenAI, and where it goes next. This is something you can't skip if you don't want to be left behind. Watch it, then read the full guide on graph engineering below.
24:33

OpenAI engineer: "At OpenAI, 85% of our engineers are running hundreds of agents with Codex. The way you do it is graph engineering." In 30-minutes lecture, he explains how a single engineer can now do the work of an entire team, how far this has already gone inside OpenAI, and where it goes next. This is something you can't skip if you don't want to be left behind. Watch it, then read the full guide on graph engineering below.

Movez

201,861 просмотров • 7 дней назад