正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

In flow matching, a coupling determines how noise and data samples are paired during training. The choice of coupling is important because it influences the geometry of trajectories at inference time. The simplest choice is the independent coupling, where noise and data points are paired arbitrarily. This can lead... show more

Alec Helbling

10,778 subscribers

65,484 次观看 • 2 个月前 •via X (Twitter)

教育科学技术

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

System identification (sysid) is the process of finding the physical parameters that make a simulation match reality. If you're training an RL locomotion policy in simulation, the accuracy of your motor model directly affects how well the policy transfers to the real robot. A recent git commit by Kevin Zakka added a sysid toolbox to MuJoCo which automates this process: you provide recorded motor data and a MuJoCo model, and it optimizes the model parameters to minimize the difference between simulated and real trajectories. For my RobStride Dynamics RS02 QDD motors (17 Nm peak, 7.75:1 gear), I built a Rust tool that sends multi-sine torque excitation at 1 kHz and records position/velocity feedback. I then feed this data into MuJoCo's sysid optimizer.

System identification (sysid) is the process of finding the physical parameters that make a simulation match reality. If you're training an RL locomotion policy in simulation, the accuracy of your motor model directly affects how well the policy transfers to the real robot. A recent git commit by Kevin Zakka added a sysid toolbox to MuJoCo which automates this process: you provide recorded motor data and a MuJoCo model, and it optimizes the model parameters to minimize the difference between simulated and real trajectories. For my RobStride Dynamics RS02 QDD motors (17 Nm peak, 7.75:1 gear), I built a Rust tool that sends multi-sine torque excitation at 1 kHz and records position/velocity feedback. I then feed this data into MuJoCo's sysid optimizer.

David Bar

48,347 次观看 • 3 个月前

Holy sh!t ! OpenAI will have their custom inference chips ready in just a few months and deployed at scale by the end of the year! 🤯 Training chip = The heavy lifters that require massive amounts of data and power to build and teach the AI models from scratch. Inference chip = The specialized, highly efficient chips that actually run the AI and generate the answers in real-time when you use it. This is going to help OpenAI drastically cut down their massive compute costs, speed up model reasoning times, and finally break free from relying entirely on Nvidia to scale their operations.

Holy sh!t ! OpenAI will have their custom inference chips ready in just a few months and deployed at scale by the end of the year! 🤯 Training chip = The heavy lifters that require massive amounts of data and power to build and teach the AI models from scratch. Inference chip = The specialized, highly efficient chips that actually run the AI and generate the answers in real-time when you use it. This is going to help OpenAI drastically cut down their massive compute costs, speed up model reasoning times, and finally break free from relying entirely on Nvidia to scale their operations.

Chris

60,278 次观看 • 4 个月前

This animation shows how a uniform grid of points is deformed by a flow matching model. The visualization highlights how the learned flow warps the underlying space—stretching, compressing, and bending it—to transform one probability distribution into another.

This animation shows how a uniform grid of points is deformed by a flow matching model. The visualization highlights how the learned flow warps the underlying space—stretching, compressing, and bending it—to transform one probability distribution into another.

Alec Helbling

42,979 次观看 • 1 年前

We’re proud to share that Tamarind Bio has been selected to build, host, and operate the inference infrastructure layer for TuneLab2.0, the next evolution of the platform. Eli Lilly and Company TuneLab is a first-of-its-kind, collaborative AI/ML drug discovery platform, bringing models trained on over $1B worth of Lilly proprietary data to the biotech ecosystem. Tamarind will power TuneLab’s scalable drug discovery workflows and model inference.

We’re proud to share that Tamarind Bio has been selected to build, host, and operate the inference infrastructure layer for TuneLab2.0, the next evolution of the platform. Eli Lilly and Company TuneLab is a first-of-its-kind, collaborative AI/ML drug discovery platform, bringing models trained on over $1B worth of Lilly proprietary data to the biotech ecosystem. Tamarind will power TuneLab’s scalable drug discovery workflows and model inference.

Deniz Kavi

43,555 次观看 • 1 个月前

The next World Cup star is playing right now, and nobody is watching. The 2026 World Cup is on, and millions of kids are dreaming of one day playing for their own flag. But the path there is almost impossible: no consolidated data, too many games, not enough scouts, human bias, and wrong place, wrong time. Any one of them can bury a generational talent before a scout ever sees them. That is what we are building to fix. So we developed a full automated annotation + action spotting model where it actually matters: grassroots football. Shaky cameras, awful angles, terrible lighting. But now nothing on the market beats sn44's v2.0 model on it. Check the last GT line, it’s human annotations. Simply because miners' outputs + post-process = magic. Post-process is a deliberate design choice. We point miners at the hardest 80% of the workflow, the part that is genuinely difficult to fake, and handle the rest downstream. Fewer places to game the system means a network you can trust. Every single challenge on sn44 can be an independent startup now. Let that sink in.

The next World Cup star is playing right now, and nobody is watching. The 2026 World Cup is on, and millions of kids are dreaming of one day playing for their own flag. But the path there is almost impossible: no consolidated data, too many games, not enough scouts, human bias, and wrong place, wrong time. Any one of them can bury a generational talent before a scout ever sees them. That is what we are building to fix. So we developed a full automated annotation + action spotting model where it actually matters: grassroots football. Shaky cameras, awful angles, terrible lighting. But now nothing on the market beats sn44's v2.0 model on it. Check the last GT line, it’s human annotations. Simply because miners' outputs + post-process = magic. Post-process is a deliberate design choice. We point miners at the hardest 80% of the workflow, the part that is genuinely difficult to fake, and handle the rest downstream. Fewer places to game the system means a network you can trust. Every single challenge on sn44 can be an independent startup now. Let that sink in.

Score - Subnet 44

430,045 次观看 • 1 个月前

Amazon’s machine learning model collects 300 million data points per season and can now predict which players are likely to blitz before the snap. This is a look into the future of broadcasting.

Amazon’s machine learning model collects 300 million data points per season and can now predict which players are likely to blitz before the snap. This is a look into the future of broadcasting.

Joe Pompliano

688,463 次观看 • 2 年前

What if you kept asking an LLM to "make it better"? In some recent work at FAIR, we investigate how we can efficiently use RL to fine-tune LLMs to iteratively self-improve on their previous solutions at inference-time. Training for iterated self-improvement can be costly. The naive approach to training for K self-improvement steps leads to K times the number of rollout steps per episode. We introduce Exploratory Iteration (ExIt), an RL-based automatic curriculum method that bootstraps diverse training distributions of self-improvement tasks by upcycling the LLM's own responses at previous turns as the starting points for both self-improvement and *self-divergence.* In order to decide what task to train on next, the curriculum prioritizes sampling of partial turn histories that led to higher return variance in its GRPO group (a learnability score that comes for free). This automatic curriculum over the bootstrapped task space teaches the model how to perform iterated self-improvement while only ever training the model on single-step self-improvement tasks. We look at ExIt's impact in both single-turn (contest math problems) and multi-turn (BFCLv3 multi-turn tasks), as well as MLE-bench, where the LLM is run in a search scaffold to produce solutions to real Kaggle competitions. Across these eval settings, we find ExIt produces models with greater capacity for inference-time self-improvement compared to GRPO. Notably, ExIt models can self-improve on test tasks for many more steps than the typical solution depth encountered during training, including a 22% improvement in MLE-bench performance compared to GRPO.

What if you kept asking an LLM to "make it better"? In some recent work at FAIR, we investigate how we can efficiently use RL to fine-tune LLMs to iteratively self-improve on their previous solutions at inference-time. Training for iterated self-improvement can be costly. The naive approach to training for K self-improvement steps leads to K times the number of rollout steps per episode. We introduce Exploratory Iteration (ExIt), an RL-based automatic curriculum method that bootstraps diverse training distributions of self-improvement tasks by upcycling the LLM's own responses at previous turns as the starting points for both self-improvement and self-divergence. In order to decide what task to train on next, the curriculum prioritizes sampling of partial turn histories that led to higher return variance in its GRPO group (a learnability score that comes for free). This automatic curriculum over the bootstrapped task space teaches the model how to perform iterated self-improvement while only ever training the model on single-step self-improvement tasks. We look at ExIt's impact in both single-turn (contest math problems) and multi-turn (BFCLv3 multi-turn tasks), as well as MLE-bench, where the LLM is run in a search scaffold to produce solutions to real Kaggle competitions. Across these eval settings, we find ExIt produces models with greater capacity for inference-time self-improvement compared to GRPO. Notably, ExIt models can self-improve on test tasks for many more steps than the typical solution depth encountered during training, including a 22% improvement in MLE-bench performance compared to GRPO.

Minqi Jiang

41,099 次观看 • 10 个月前

OpenAI's Deep Research is getting a run for its money. Deep Lake was just released, and it's a different take on an AI system that can do deep research on your own data. You can use Deep Lake to build AI search with reasoning on your private and public data. (Look at the attached videos to get an idea of how it works.) If you want to research proprietary and sensitive data, Deep Research won't help you because it's limited to public data. Deep Lake, however, will allow you to use your private data. On top of that, Deep Lake supports multi-modal retrieval from the ground up. It uses vision language models for data ingestion and retrieval so that you can connect any data (PDFs, images, videos, structured data, etc.) You can even use mixed-data queries! Deep Lake can search your data from S3, Dropbox, and GCP. It learns from your queries over time, making the results as relevant to your work as possible!

OpenAI's Deep Research is getting a run for its money. Deep Lake was just released, and it's a different take on an AI system that can do deep research on your own data. You can use Deep Lake to build AI search with reasoning on your private and public data. (Look at the attached videos to get an idea of how it works.) If you want to research proprietary and sensitive data, Deep Research won't help you because it's limited to public data. Deep Lake, however, will allow you to use your private data. On top of that, Deep Lake supports multi-modal retrieval from the ground up. It uses vision language models for data ingestion and retrieval so that you can connect any data (PDFs, images, videos, structured data, etc.) You can even use mixed-data queries! Deep Lake can search your data from S3, Dropbox, and GCP. It learns from your queries over time, making the results as relevant to your work as possible!

Santiago

171,340 次观看 • 1 年前

BIG NEWS. The Blockworks website has evolved: yesterday we were the home of news, today we are the home of onchain data. Head over to the site to see for yourself, but here's a little snippet of what you can expect: 1. Sector leaderboards (chains, DEXes, borrow lend, DATs, etc...) 2. Comprehensive data dashboards protocols 3. The ability to compare pricing and onchain data easily (coming soon) We're doing this because the industry still has a gigantic data problem. As investors get more sophisticated and fundamentals driven, basic high level facts are no longer sufficient. Investors need to be able to trust the data they are seeing and go much deeper than the surface level info that's available today. Additionally, because many data providers allow companies to essentially self report, you can't trust what you are seeing. This site is our contribution to fixing that problem and to ensuring clear, accurate data for investors. Blockworks is fully dedicated to becoming the most comprehensive data company in crypto in 2026. This is the first of many, many announcements like this this year, stay tuned.

BIG NEWS. The Blockworks website has evolved: yesterday we were the home of news, today we are the home of onchain data. Head over to the site to see for yourself, but here's a little snippet of what you can expect: 1. Sector leaderboards (chains, DEXes, borrow lend, DATs, etc...) 2. Comprehensive data dashboards protocols 3. The ability to compare pricing and onchain data easily (coming soon) We're doing this because the industry still has a gigantic data problem. As investors get more sophisticated and fundamentals driven, basic high level facts are no longer sufficient. Investors need to be able to trust the data they are seeing and go much deeper than the surface level info that's available today. Additionally, because many data providers allow companies to essentially self report, you can't trust what you are seeing. This site is our contribution to fixing that problem and to ensuring clear, accurate data for investors. Blockworks is fully dedicated to becoming the most comprehensive data company in crypto in 2026. This is the first of many, many announcements like this this year, stay tuned.

Mippo 🟪

91,821 次观看 • 6 个月前

This is the level of noise in Bibu! Eileen on the gate has a unique charm and screams like a wookie! Our vets are checking Klaas as he had fever this morning. Its very important to act on time because they can get infections quickly due to their open belly button! #アザラシ幼稚園

This is the level of noise in Bibu! Eileen on the gate has a unique charm and screams like a wookie! Our vets are checking Klaas as he had fever this morning. Its very important to act on time because they can get infections quickly due to their open belly button! #アザラシ幼稚園

Panka

46,860 次观看 • 1 年前

Fancy winning a pair of #EFL tickets for a home game of your choice? 😍 Reply with the answer to the following question and use #Super6Christmas for a chance to win 🎟️ How many points are Leicester City on in the league this season? 🦊 Winner announced at 4pm, T&Cs apply ⌚️

Fancy winning a pair of #EFL tickets for a home game of your choice? 😍 Reply with the answer to the following question and use #Super6Christmas for a chance to win 🎟️ How many points are Leicester City on in the league this season? 🦊 Winner announced at 4pm, T&Cs apply ⌚️

Super 6

53,495 次观看 • 2 年前

AI Is Moving Beyond “Generating Videos” — Toward “Generating Worlds” Over the past two years, AI video models have advanced at an astonishing pace. From Runway and Pika to Sora and Veo, AI-generated videos have become increasingly realistic and more consistent with the physical laws of the real world. Many people believe the next objective is simply to generate videos that are longer, sharper, and more lifelike. But if we take a step back, we can see that the real transformation is not happening in video itself. It is happening in world models. What Is a World Model? In 1943, psychologist Kenneth Craik proposed an idea that would influence artificial intelligence research for decades. He argued that the human brain does not merely react to the outside world. Instead, it maintains an internal model of how the world works. Because we have this internal model, we can predict the outcome of an action before we actually take it. Before crossing a road, we estimate whether a car will pass by. Before catching a ball, we predict its trajectory. These abilities come from continuously simulating the world in our minds, rather than relying entirely on trial and error. This idea later became known by a more formal term: World Model. A world model does not describe a single image or a fixed video clip. It is an internal representation capable of continuously simulating the rules and dynamics of the real world. Why Is AI Research Turning Toward World Models? Because predicting “what comes next” is becoming increasingly central to how AI systems work. Language models predict the next token. Image models predict the next step in the denoising process. Video models predict the next frame. A world model, however, attempts to predict something broader: What should the world look like in the next moment? In 2018, David Ha and Jürgen Schmidhuber proposed in their paper World Models that an intelligent agent could first learn a model of the world, and then use that internal model to plan its actions. The Dreamer series later demonstrated that many complex tasks could be learned by training agents inside an “imagined world.” At the same time, the development of video models such as Sora and Veo led researchers to another realization: A model capable of continuously generating video has already learned, at least implicitly, many of the rules governing the real world. As a result, these two research directions have gradually begun to converge. But Video Is Not Yet a World This is where the distinction is often misunderstood. For a world model to support meaningful real-time interaction, it must solve several critical problems. Most video models today are essentially answering one question: What should the next frame look like? A true world model needs to answer much more: What happens if I take one step forward? If I walk behind a building and then return, will the building still be there? If I suddenly change the camera angle, will the entire space remain consistent? If I enter a command such as: “Summon a dragon.” Will the world respond immediately? In other words, a world model must do more than generate content. It must understand space. It must understand time. It must understand causality. And it must understand interaction. Moving from watching to participating is where the real difficulty of world models begins. World Models Are Entering the Interactive Era One of the latest attempts in this direction is Alaya World, recently open-sourced by Alaya World, or Alaya Lab. Instead of generating a fixed video clip, it generates a world that users can explore in real time. Users can begin with text, an image, or a video, enter the generated scene, move freely through it, and introduce new prompts at any moment during generation. The world responds immediately. According to the publicly released information, Alaya World provides: Real-time streaming generation at 720p and 24 FPS Stable continuous exploration for more than one minute The ability to switch prompts and trigger skills or events during generation Model weights and inference code released under the Apache 2.0 License Training code and datasets planned for future release What makes these capabilities important is not simply the technical specifications. It is that the generated “world” can now support continuous interaction. The official demo shows that users can genuinely control, transform, and explore the generated environment. AI Is Evolving From a Tool Into an Environment Over the past few years, most discussions around AI have focused on content generation. Generating text. Generating images. Generating videos. But world models raise a fundamentally different question: Can AI generate an environment that people can inhabit, explore, and continuously evolve? If the answer is yes, the impact will extend far beyond video generation. Game development, robotics training, embodied intelligence, digital twins, virtual production, and many other fields could be transformed by the development of world models. World models are still at a very early stage. Yet from Craik’s proposal of an internal mental model more than eighty years ago to the emergence of today’s interactive world-generation systems, a clear evolutionary path is beginning to take shape. Perhaps what AI is ultimately learning has never been limited to images, videos, or language. Perhaps it is learning the world itself. References GitHub: Technical Report:

AI Is Moving Beyond “Generating Videos” — Toward “Generating Worlds” Over the past two years, AI video models have advanced at an astonishing pace. From Runway and Pika to Sora and Veo, AI-generated videos have become increasingly realistic and more consistent with the physical laws of the real world. Many people believe the next objective is simply to generate videos that are longer, sharper, and more lifelike. But if we take a step back, we can see that the real transformation is not happening in video itself. It is happening in world models. What Is a World Model? In 1943, psychologist Kenneth Craik proposed an idea that would influence artificial intelligence research for decades. He argued that the human brain does not merely react to the outside world. Instead, it maintains an internal model of how the world works. Because we have this internal model, we can predict the outcome of an action before we actually take it. Before crossing a road, we estimate whether a car will pass by. Before catching a ball, we predict its trajectory. These abilities come from continuously simulating the world in our minds, rather than relying entirely on trial and error. This idea later became known by a more formal term: World Model. A world model does not describe a single image or a fixed video clip. It is an internal representation capable of continuously simulating the rules and dynamics of the real world. Why Is AI Research Turning Toward World Models? Because predicting “what comes next” is becoming increasingly central to how AI systems work. Language models predict the next token. Image models predict the next step in the denoising process. Video models predict the next frame. A world model, however, attempts to predict something broader: What should the world look like in the next moment? In 2018, David Ha and Jürgen Schmidhuber proposed in their paper World Models that an intelligent agent could first learn a model of the world, and then use that internal model to plan its actions. The Dreamer series later demonstrated that many complex tasks could be learned by training agents inside an “imagined world.” At the same time, the development of video models such as Sora and Veo led researchers to another realization: A model capable of continuously generating video has already learned, at least implicitly, many of the rules governing the real world. As a result, these two research directions have gradually begun to converge. But Video Is Not Yet a World This is where the distinction is often misunderstood. For a world model to support meaningful real-time interaction, it must solve several critical problems. Most video models today are essentially answering one question: What should the next frame look like? A true world model needs to answer much more: What happens if I take one step forward? If I walk behind a building and then return, will the building still be there? If I suddenly change the camera angle, will the entire space remain consistent? If I enter a command such as: “Summon a dragon.” Will the world respond immediately? In other words, a world model must do more than generate content. It must understand space. It must understand time. It must understand causality. And it must understand interaction. Moving from watching to participating is where the real difficulty of world models begins. World Models Are Entering the Interactive Era One of the latest attempts in this direction is Alaya World, recently open-sourced by Alaya World, or Alaya Lab. Instead of generating a fixed video clip, it generates a world that users can explore in real time. Users can begin with text, an image, or a video, enter the generated scene, move freely through it, and introduce new prompts at any moment during generation. The world responds immediately. According to the publicly released information, Alaya World provides: Real-time streaming generation at 720p and 24 FPS Stable continuous exploration for more than one minute The ability to switch prompts and trigger skills or events during generation Model weights and inference code released under the Apache 2.0 License Training code and datasets planned for future release What makes these capabilities important is not simply the technical specifications. It is that the generated “world” can now support continuous interaction. The official demo shows that users can genuinely control, transform, and explore the generated environment. AI Is Evolving From a Tool Into an Environment Over the past few years, most discussions around AI have focused on content generation. Generating text. Generating images. Generating videos. But world models raise a fundamentally different question: Can AI generate an environment that people can inhabit, explore, and continuously evolve? If the answer is yes, the impact will extend far beyond video generation. Game development, robotics training, embodied intelligence, digital twins, virtual production, and many other fields could be transformed by the development of world models. World models are still at a very early stage. Yet from Craik’s proposal of an internal mental model more than eighty years ago to the emergence of today’s interactive world-generation systems, a clear evolutionary path is beginning to take shape. Perhaps what AI is ultimately learning has never been limited to images, videos, or language. Perhaps it is learning the world itself. References GitHub: Technical Report:

雪踏乌云

112,114 次观看 • 17 天前

At Dreamforce, the exploration of the possibilities of AI is incredible. At the same time, it's important that we understand that policies never move at the same speed as technology. So having a grasp over what is happening and how the negative aspects of AI don't proliferate in a big way, to minimize collateral damage is most important. I hope the companies and governments across the world wake up to this and ensure the best of AI possibilities will reach the people, as this is a phenomenal enabler for humanity like never before. -Sg Marc Benioff Dreamforce Salesforce #DF24

At Dreamforce, the exploration of the possibilities of AI is incredible. At the same time, it's important that we understand that policies never move at the same speed as technology. So having a grasp over what is happening and how the negative aspects of AI don't proliferate in a big way, to minimize collateral damage is most important. I hope the companies and governments across the world wake up to this and ensure the best of AI possibilities will reach the people, as this is a phenomenal enabler for humanity like never before. -Sg Marc Benioff Dreamforce Salesforce #DF24

Sadhguru

68,302 次观看 • 1 年前

Depth Any Video with Scalable Synthetic Data AI physicists and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.

Depth Any Video with Scalable Synthetic Data AI physicists and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.

MrNeRF

27,428 次观看 • 1 年前

Why is the medical community staying silent? I asked if other providers are reporting vaccine injuries to VAERS, and the answer was a staggering "no." If we aren’t tracking the data, how can we ensure patient safety? With over 19,500 deaths reported in the system, it's time to start asking why the burden of reporting is falling on so few.

Why is the medical community staying silent? I asked if other providers are reporting vaccine injuries to VAERS, and the answer was a staggering "no." If we aren’t tracking the data, how can we ensure patient safety? With over 19,500 deaths reported in the system, it's time to start asking why the burden of reporting is falling on so few.

Mary Talley Bowden MD

26,182 次观看 • 5 个月前

Day 11/90 of Inference Engineering How does vLLM work and how is it used in production? Before we discuss how vLLM works internally, it helps to understand what vLLM is. At a high level, vLLM is an inference engine that is designed to serve LLMs to thousands of concurrent users efficiently while managing scarce compute and memory. The goal for vLLM is to maximize throughput and minimize latency; optimizing for the best inference economics and experience for end users. With every request from the end user, it eventually ends up in the engine core, gets scheduled alongside other requests from other concurrent users, executes on the GPU, and updates the KV cache with the new key and value vectors, and streams the tokens back to the user. The Scheduler decides what requests should execute next while continuously batching requests together to maximize GPU utilization. Continuous batching is an inference optimization that allows new requests to join a running batch as other requests finish generating tokens. This helps with keeping the GPU utilization high instead of letting it sit idle waiting for an entire batch to complete generating. After the scheduler dispatches the selected batch to the Model Executor, the Model Executor prepares the tensors and metadata required for inference, retrieves each request’s block table from KV Cache Manager, launches the optimized transformer forward pass on the GPU, computes the logits, updates the KV cache with the new key and value vectors, and finally returns the results for sampling and streaming. The KV Cache Manager uses the PagedAttention memory layout to allocate fixed-size cache blocks on demand and maintains a Free Block Queue on the CPU that tracks which blocks in the GPU’s Paged KV Cache are currently free. When a request needs additional KV cache space, the KV Cache manager takes a free block from the queue and assigns it to that request, thus avoiding an expensive search through GPU memory for available cache blocks. All of these components form the core of vLLM’s inference engine. The Scheduler determines what requests are executed, the Model Executor determines how those requests are executed, the KV Cache Manager determines where each request’s KV cache lives using the PagedAttention Memory Layout. This architecture enables vLLM to serve thousands of concurrent requests with high throughput, low latency, and efficient GPU memory utilization. Heres a little animation that visualizes everything! - I've also completed the forward pass for my mnist.c project. I had a nice chat with shrey birmiwal, such a knowledgeable guy. Excited to learn more about vLLM and implement a tiny-vLLM one day.

Day 11/90 of Inference Engineering How does vLLM work and how is it used in production? Before we discuss how vLLM works internally, it helps to understand what vLLM is. At a high level, vLLM is an inference engine that is designed to serve LLMs to thousands of concurrent users efficiently while managing scarce compute and memory. The goal for vLLM is to maximize throughput and minimize latency; optimizing for the best inference economics and experience for end users. With every request from the end user, it eventually ends up in the engine core, gets scheduled alongside other requests from other concurrent users, executes on the GPU, and updates the KV cache with the new key and value vectors, and streams the tokens back to the user. The Scheduler decides what requests should execute next while continuously batching requests together to maximize GPU utilization. Continuous batching is an inference optimization that allows new requests to join a running batch as other requests finish generating tokens. This helps with keeping the GPU utilization high instead of letting it sit idle waiting for an entire batch to complete generating. After the scheduler dispatches the selected batch to the Model Executor, the Model Executor prepares the tensors and metadata required for inference, retrieves each request’s block table from KV Cache Manager, launches the optimized transformer forward pass on the GPU, computes the logits, updates the KV cache with the new key and value vectors, and finally returns the results for sampling and streaming. The KV Cache Manager uses the PagedAttention memory layout to allocate fixed-size cache blocks on demand and maintains a Free Block Queue on the CPU that tracks which blocks in the GPU’s Paged KV Cache are currently free. When a request needs additional KV cache space, the KV Cache manager takes a free block from the queue and assigns it to that request, thus avoiding an expensive search through GPU memory for available cache blocks. All of these components form the core of vLLM’s inference engine. The Scheduler determines what requests are executed, the Model Executor determines how those requests are executed, the KV Cache Manager determines where each request’s KV cache lives using the PagedAttention Memory Layout. This architecture enables vLLM to serve thousands of concurrent requests with high throughput, low latency, and efficient GPU memory utilization. Heres a little animation that visualizes everything! - I've also completed the forward pass for my mnist.c project. I had a nice chat with shrey birmiwal, such a knowledgeable guy. Excited to learn more about vLLM and implement a tiny-vLLM one day.

max fu

70,024 次观看 • 16 天前

🚨 BREAKING: Gravity doesn’t actually “pull” you. That’s just how it looks. What’s really happening is deeper… Space-time isn’t empty. It behaves more like a field with structure. And when mass appears… it doesn’t just “bend space” it changes how time flows. Think of it like this: If time flows evenly → everything stays stable If time compresses → motion is forced toward that region That motion is what we call gravity. In Einstein’s model: mass curves space-time But here’s the missing piece: What is doing the curving? In my framework: Mass = a region where time is compressed Gravity = objects moving along time gradients Motion = the system trying to rebalance So you’re not being “pulled down”… You’re being carried along a flow of time. This explains something most people miss: Gravity always points one way. Because time only flows one way. And the deeper implication is wild: If gravity is just time imbalance… then controlling time flow = controlling gravity. The real question is: Are we discovering gravity… or learning how to engineer it? Follow for more this goes deeper.

🚨 BREAKING: Gravity doesn’t actually “pull” you. That’s just how it looks. What’s really happening is deeper… Space-time isn’t empty. It behaves more like a field with structure. And when mass appears… it doesn’t just “bend space” it changes how time flows. Think of it like this: If time flows evenly → everything stays stable If time compresses → motion is forced toward that region That motion is what we call gravity. In Einstein’s model: mass curves space-time But here’s the missing piece: What is doing the curving? In my framework: Mass = a region where time is compressed Gravity = objects moving along time gradients Motion = the system trying to rebalance So you’re not being “pulled down”… You’re being carried along a flow of time. This explains something most people miss: Gravity always points one way. Because time only flows one way. And the deeper implication is wild: If gravity is just time imbalance… then controlling time flow = controlling gravity. The real question is: Are we discovering gravity… or learning how to engineer it? Follow for more this goes deeper.

TheNewPhysics

21,772 次观看 • 3 个月前

Contrail lesson! 1. “Chemtrails” don’t exist. Just to get that out of the way. 2. Observe the satellite loop and Skew-T chart. In the IR satellite loop you can see yesterday, the West Coast had a decent short wave ridge suppressing moisture over California and Nevada. Today, you can see moisture from a low pressure over the Pacific spilling over the ridge that is now moving east of California. This is upper level moisture ADVECTING into the area. This upper level moisture is mainly above the 500mb level, or 20,000ft. 3. Now observe the Skew-T chart. Particularly clue into the 300mb level. This is a perfect example of what I talk about all the time, and why it’s important to pay attention to the 300mb level. This moisture layer is advecting particularly at the 300mb level, and synoptic scale cirrus development, and advection, typically occurs at 300mb. This is key because aircraft are flying at and above the 300mb level. 4. So, lastly, observe the pictures that I took of the sky over northern Nevada at the time of this post. You can see the layer of cirrus as well as contrails persisting in that moisture layer, exactly as depicted in the satellite shot AND confirmed by the Skew-T chart. Keep in mind that temperatures at this level of the atmosphere are typically -20 to -50°C. In this case, you can see that the temperature at 300mb is -40°C and relative humidities at this level are far different than what you experience at the surface. Any decrease in the gap between temperature and dewpoint at this level can significantly increase the relative humidity. This is why it’s referred to as “relative”because it’s far different than temperatures and dew points at the surface. So, to bring it all together, aircraft flying at these altitudes, which most commercial and military aircraft do, injecting warm, moist air from the engines rapidly into the super cooled environment, not only instantly form contrails, but when relative humidities are as depicted in this example, will enable contrails to persist for hours at a time supported by the moisture existing in that layer. This is what causes persistent contrails. These ARE NOT “chemtrails” and because they persist, does not, and will not ever, make them “chemtrails.” Now that you all needed your government to tell you that climate change was a hoax and I’ve been telling you for years that the “Geoengineering” and “chemtrail” nonsense are propaganda directly related to the climate change hoax, hopefully you can take some time to learn the basics of the atmosphere and understand what I’m showing you here, and how it works, so you’re not fooled by climate propaganda going forward. Thank you for your attention to this matter. 💪🏼🇺🇸

Contrail lesson! 1. “Chemtrails” don’t exist. Just to get that out of the way. 2. Observe the satellite loop and Skew-T chart. In the IR satellite loop you can see yesterday, the West Coast had a decent short wave ridge suppressing moisture over California and Nevada. Today, you can see moisture from a low pressure over the Pacific spilling over the ridge that is now moving east of California. This is upper level moisture ADVECTING into the area. This upper level moisture is mainly above the 500mb level, or 20,000ft. 3. Now observe the Skew-T chart. Particularly clue into the 300mb level. This is a perfect example of what I talk about all the time, and why it’s important to pay attention to the 300mb level. This moisture layer is advecting particularly at the 300mb level, and synoptic scale cirrus development, and advection, typically occurs at 300mb. This is key because aircraft are flying at and above the 300mb level. 4. So, lastly, observe the pictures that I took of the sky over northern Nevada at the time of this post. You can see the layer of cirrus as well as contrails persisting in that moisture layer, exactly as depicted in the satellite shot AND confirmed by the Skew-T chart. Keep in mind that temperatures at this level of the atmosphere are typically -20 to -50°C. In this case, you can see that the temperature at 300mb is -40°C and relative humidities at this level are far different than what you experience at the surface. Any decrease in the gap between temperature and dewpoint at this level can significantly increase the relative humidity. This is why it’s referred to as “relative”because it’s far different than temperatures and dew points at the surface. So, to bring it all together, aircraft flying at these altitudes, which most commercial and military aircraft do, injecting warm, moist air from the engines rapidly into the super cooled environment, not only instantly form contrails, but when relative humidities are as depicted in this example, will enable contrails to persist for hours at a time supported by the moisture existing in that layer. This is what causes persistent contrails. These ARE NOT “chemtrails” and because they persist, does not, and will not ever, make them “chemtrails.” Now that you all needed your government to tell you that climate change was a hoax and I’ve been telling you for years that the “Geoengineering” and “chemtrail” nonsense are propaganda directly related to the climate change hoax, hopefully you can take some time to learn the basics of the atmosphere and understand what I’m showing you here, and how it works, so you’re not fooled by climate propaganda going forward. Thank you for your attention to this matter. 💪🏼🇺🇸

Dylan Tucker

26,804 次观看 • 9 个月前

Model-Free Reinforcement Learning (MFRL) has been alluring, especially with supercharged compute with physics on GPU. However, the methods use 0-th order gradients, and are often not the best optimizers. Can we do better than PPO in continuous control for robotics? Turns out yes! 🥳 tl;dr: Faster, better RL than PPO in continuous control 💪 The answer lies in using more information from the simulation. We are juicing the simulation on GPU as it is, why not use it for gradients as well? This has been a driving question in a series of our works. We first studied this problem in ICLR 2022 paper on Short Horizon Actor Critic Naive gradient based methods are stuck in local minima and have exploding/vanishing gradients. SHAC solved this problem truncated rollouts and model based value estimation, where the model is Differentiable Sim. This boosted sample efficiency and wall-clock time immensely especially in high dimensional systems such as humanoids Yet, given enough compute PPO often caught up. Our follow up paper on on Adaptive Horizon Actor Critic at ICML 2024 discovers the cause and provides a fix. However, we find that even when given ground-truth dynamics, not all gradients are useful due to sample error. 1st-Order Model-Based Reinforcement Learning methods employing differentiable simulation provide gradients with reduced variance but are susceptible to bias in scenarios involving stiff dynamics, such as physical contact. We find that back-propagating through contact and long trajectories drastically reduces gradient accuracy. Using this insight, we propose AHAC to dynamically adapt its roll-out horizon to avoid differentiating through stiff contact. AHAC is a first-order model-based RL algorithm that learns high-dimensional tasks in minutes (wall clock) and outperforms PPO by 40%, even in the limit of data provided to PPO. This work is led by Ignat Georgiev alongside Krishnan Srinivasan, Jie Xu, Eric Heiden and ample assistance from warp team at NVIDIA Robotics (Miles Macklin)

Model-Free Reinforcement Learning (MFRL) has been alluring, especially with supercharged compute with physics on GPU. However, the methods use 0-th order gradients, and are often not the best optimizers. Can we do better than PPO in continuous control for robotics? Turns out yes! 🥳 tl;dr: Faster, better RL than PPO in continuous control 💪 The answer lies in using more information from the simulation. We are juicing the simulation on GPU as it is, why not use it for gradients as well? This has been a driving question in a series of our works. We first studied this problem in ICLR 2022 paper on Short Horizon Actor Critic Naive gradient based methods are stuck in local minima and have exploding/vanishing gradients. SHAC solved this problem truncated rollouts and model based value estimation, where the model is Differentiable Sim. This boosted sample efficiency and wall-clock time immensely especially in high dimensional systems such as humanoids Yet, given enough compute PPO often caught up. Our follow up paper on on Adaptive Horizon Actor Critic at ICML 2024 discovers the cause and provides a fix. However, we find that even when given ground-truth dynamics, not all gradients are useful due to sample error. 1st-Order Model-Based Reinforcement Learning methods employing differentiable simulation provide gradients with reduced variance but are susceptible to bias in scenarios involving stiff dynamics, such as physical contact. We find that back-propagating through contact and long trajectories drastically reduces gradient accuracy. Using this insight, we propose AHAC to dynamically adapt its roll-out horizon to avoid differentiating through stiff contact. AHAC is a first-order model-based RL algorithm that learns high-dimensional tasks in minutes (wall clock) and outperforms PPO by 40%, even in the limit of data provided to PPO. This work is led by Ignat Georgiev alongside Krishnan Srinivasan, Jie Xu, Eric Heiden and ample assistance from warp team at NVIDIA Robotics (Miles Macklin)

Animesh Garg

52,300 次观看 • 2 年前

I'm generally curious how you can have two 'truths' at the same time. Michelle Wu claims that: 1. Boston is the safest major city in the country, safe for everyone, and illegal immigrants do not contribute to the crime. Areas are not 'less safe' because of this. 2. Boston law enforcement does not collect immigration status during arrests. So my question is, if law enforcement does not track or collect immigration status, as she claims, how can the city confidently claim that illegal immigrants are not contributing to crime without the data?

I'm generally curious how you can have two 'truths' at the same time. Michelle Wu claims that: 1. Boston is the safest major city in the country, safe for everyone, and illegal immigrants do not contribute to the crime. Areas are not 'less safe' because of this. 2. Boston law enforcement does not collect immigration status during arrests. So my question is, if law enforcement does not track or collect immigration status, as she claims, how can the city confidently claim that illegal immigrants are not contributing to crime without the data?

Mike Urban

26,225 次观看 • 1 年前