Compute & System Design for Next Generation Frontier Models

Talk transcript

1. 0:00

   [on-hold electronic music] Couple different things, right? Like, you know, people have been talking about stagnation, um, and, uh, it's-- I, I don't think anyone else, anyone here sees that.

2. 0:21

   But a lot of people have been talking about stagnation of models, and a lot of the, a lot of that has to just do with the fact that we haven't seen a big capabilities leap, uh, in the last bit.

3. 0:31

   Uh, but that, that comes really from, uh, models that we're using today are largely the same as the models that were trained in 2022, right? GPT-4, 4 Turbo, 4o, those are just smaller models that are trained for longer, so similar quality, right?

4. 0:45

   Um, you know, 3.5 Sonnet came out recently, but again, that's actually smaller than Opus, but it's somehow better because they trained it for longer, right? But we haven't seen a extremely large model come out yet and, and but we will soon.

5. 0:57

   Uh, but one interesting thing, right, is GPT-4 is like one point eight trillion parameters. It's crazy, crazy expensive to run, right? Uh, two hundred billion parameters. Uh, e-each token requires, you know, fif-- almost six hundred gigaflops.

6. 1:10

   Uh, but that, that, that's almost gonna be considered a last generation model, right, in, in a year from now. Um, so there's a couple of things that I wanted to talk about regarding that, right?

7. 1:19

   And, and mostly on the inference side, because I don't think, you know, anyone here is gonna try and train that kind of next generation model, but definitely we're-- we need to be able to run it.

8. 1:27

   Um, and so, you know, a few things, right? So just, just gonna break down inference, uh, in detail, right? Uh, you know, uh, you know, there's two parts of inference, right?

9. 1:36

   There's prefill, there's decode. Prefill is the prompt processing, right? And the interesting thing is if you have a two K prompt, two K, uh, context length prompt, right, two thousand tokens you input into GPT, um, that's, that's a petaflop itself, right?

10. 1:50

    Um, and then, you know, if you have thirty-two thousand prompt that you enter, it's twenty petaflops actually. So it's, uh, an incredible amount of compute, uh, that's required to just process the prompt.

11. 2:00

    Um, and, and, you know, while, while prefill is, is very compute intensive, right, it's actually the opposite of decode, right? Decode is actually generating each token iteratively, right? So you, you process the prompt, then you ge-generate a token, you feed it back in, and you keep going iteratively.

12. 2:16

    Right? Um, and decode is extremely memory bandwidth intensive, right? Um, you have to load the whole model from the weights, the, the entire-- all the weights into the, uh, chip, right, or chips, uh, for decode.

13. 2:29

    Um, and the big challenge here is that, you know, hey, if you have one point eight trillion parameters, if you're running at a reasonable batch size, you're activating all the experts, you need to re-- load all one point eight trillion parameters every single token generation, right?

14. 2:44

    Even if you're serving multiple users at once, that means you're, uh, you, you need, you know, a one point eight, uh, you need terabytes a second of memory bandwidth.

15. 2:52

    You wanna do thirty tokens per second. I think that's like a minimum bar for most people, right? Uh, a lot of people want hundreds of tokens per second. But even if you want thirty tokens per second per user, sixty-four users, you need sixty terabytes a second of memory bandwidth.

16. 3:05

    Uh, if you, if-- Even if you look at an H100, it has like three, right? So this is a extremely challenging systems problem. Um, more, you know, decode, while it is very bandwidth intensive, it's actually quite cheap on the compute, which is why, uh, if you look at like OpenAI pricing or Claude pricing, you see a three

17. 3:22

    or four-to-one ratio between prefill versus decode pricing, right? Uh, so the input tokens cost, you know, one-third that of the output tokens, um, or one-fourth that. So, so, you know, today the best models I think, uh, 4o and, and 3.5 Sonnet are, uh, I wanna say it's fifteen dollars per million tokens, and then it's five dollars per

18. 3:43

    million tokens for input, uh, fifteen for output. Um, so five for prefill, fifteen for decode. Um, and, and soon we're gonna have, you know, in the, in the open source, you know, so what everyone here can touch is, is Llama 3, 405B, right?

19. 3:56

    And that's, that's gonna be a real capability sort of unlock, um, for the, you know, the open source market as well as, you know, builders here, right? And I think, I think there's a couple things that, uh, people really need to be able to implement, right?

20. 4:10

    Like, you can't just run llama.cpp on Llama 405B, right? Like, it's just not gonna work. Um, so there's a bunch of stuff that people have to work on, um, you know, whether it's using, you know, closed source libraries like TensorRT-LLM, uh, that only work on NVIDIA or like vLLM, which is an open source library that works, uh,

21. 4:28

    on AMD and Intel and, and soon other people's chips as well. Um, you know, there's, there's a lot of stuff that people need to figure out. One, one of those is, is continuous batching, right?

22. 4:37

    Uh, 'cause you're gonna get, you know, running inference at batch size one is horrendously expensive. Um, you know, it's great to run it on-- if you're running it on your own personal devices.

23. 4:46

    But if you're running it in the cloud, right, you're renting GPUs, you're running batch size one, you're, you're gonna cost yourself ten X more. You know, ten X is a low bar, right?

24. 4:55

    It's actually could be ten X to a hundred X more than running at a high batch, right? So you have to figure out how to run high batch sizes.

25. 5:01

    Batch size is how many concurrent users you're serving. Um, and so one of those things that makes it difficult is that users' requests come in at different times, right?

26. 5:10

    Uh, one person might send a request now, and then another person sends in a request five seconds later, uh, but the first person's request is not done. So you need to be able to do continuous batching, i.e., sub-- uh, be able to run through the model iteratively, uh, every time, right?

27. 5:23

    Um, and, and bring in new users. So continuous batching is one of the things that you have to have to have support of, and, and a lot of software today like llama.cpp doesn't have support for that.

28. 5:32

    So either you need to build it yourself or, um, you know, contribute to an open source project that, that builds this, um, to, to enable

29. 5:41

    low cost inference, right, for, you know, models like Llama 405B, right? Um, another one of those is, is, uh, disaggregated, uh, prefill or disaggregated batching, right? Depends on what you call it.

30. 5:54

    Um, but you know, if you go back to earlier, I was discussing, uh, prefill is very compute-intensive, decode is very, uh, bandwidth-intensive. These are two different workloads, but when you ser-- when you're serving a user, right, whether it's, uh, you know, in your own app or you're using an API, what have you, right?

31. 6:10

    Like, these users, uh, don't care that it's two different workloads, right? It's one workload to them. Uh, I get tokens out, right? I submit something to you, and I get tokens back.

32. 6:19

    Uh, but, but for anyone running the infra themselves, uh, they need to, they need to be keenly aware that these are two different workloads. Um, so one thing that a lot of people have, uh, started to do, um, Google's publicly said they're doing it, I believe OpenAI and Anthropic are also doing it.

33. 6:33

    Um, you know, uh, other firms like Together and Fireworks have hinted that they're doing this, uh, is, is disaggregated prefill, right? So once your inference volumes are high enough, you don't just run inference...

34. 6:46

    You know, you don't just replicate the model across however many chips you have, right? Uh, say, say it takes four model-- four chips to serve Llama 405B, right, in the future.

35. 6:54

    Um, you wouldn't just ha-- You know, if you have so many-- if you have enough users, you don't just go four and then eight, 16, whatever, right? You don't just replicate that across the world.

36. 7:03

    You actually do this thing called disaggregated prefill. You have one set of accelerators do the prefill, which is very compute-intensive, and then you hand it off to the other set of accelerators to do decode.

37. 7:13

    Now, today, everyone just uses the same accelerator for that, right? H100 or A100 or, you know, maybe, maybe L40 or something, but mostly H100. Um,

38. 7:23

    but there's a, there's a reason you do this, right? And, and, and that big reason is that you have a lot of noisy neighbors, right? Um, so if you've ever worked in like CPUs or on anything in cloud computing, noisy neighbors are a huge, huge issue.

39. 7:35

    Um, and actually, like, there's-- it's very trivial to dramatically slow down most inference providers' services, uh, if you, if you just, uh, send queries in a certain way, like in a, in a sort of malicious way.

40. 7:46

    Um, you can, you can just slow down people's, uh,

41. 7:50

    service, right? Whether that's, you know... And, and that'll, that'll impact the user's time to first token, right? Um, and I think that's a huge issue, right? If time to first token is too long, people will just quit, right, using your service.

42. 8:01

    Um, if, uh, you know, the tokens per second varies a lot, right? For a moment, you're getting a hundred tokens per second, and then it drops down to like 30, then it drops-- it goes back up to a hundred.

43. 8:11

    That's gonna be really annoying to the user. So, so there's a lot of things around, you know, SLA and, and reliability and all these things that you have to guarantee.

44. 8:19

    And so disaggregated prefill, uh, is, is one of the techniques to do that, right? Um, and, and so you don't wanna have someone submit, you know, for example, "Hey, I have a database, and I wanna sub-- I wanna run an LLM query across every single row in that database, and I'm just gonna submit it to you, my

45. 8:36

    service provider, because you have this cool model or what have you that's fine-tuned on some data set." And, well, whatever it is, right? If I submit ten thousand rows to you at once, that's gonna kill everyone else's performance, right?

46. 8:47

    So, so this is one of the techniques that people have for, uh, making it so, you know, that, that person who you definitely wanna serve, uh, doesn't impact everyone else's usage.

47. 8:57

    Uh, because once you open up your service to the real world, you're not gonna be able to control who's submitting what, and rate limits are the most annoying thing ever, so that's not the correct way to go about it.

48. 9:06

    Um, another thing is context caching, right? So Google launched this recently. Uh, they're the only one offering this today, but I think this is a really big deal, uh, 'cause when people talk about fine-tuning, right, of models, that's great.

49. 9:19

    Uh, but in reality, the best models are really expensive to fine-tune or impossible to fine-tune, right? I can't go fine-tune 3.5 Sonnet, or fine-tuning Llama 405B is gonna take, you know, dozens and dozens of GPUs, right?

50. 9:33

    So, so instead of that, the, the, uh... Or, you know, and, and in closed source models generally. So Google only does closed source models mostly for the big ones, right?

51. 9:40

    So Gemini 1.5 Pro, they offered this-- they, they brought this recently, right? Which is context caching. So instead of, you know, fine-tuning your model, why not, you know, just fill out a context length of...

52. 9:51

    You know, they, they offer, I think, two million now today, right? Two million context length. Um, why not fill it out with your data there, right? Um, you know, and, and there's a couple, you know, advantages to that.

53. 10:02

    One is you can use the best models, right? In the case of fine-tuned models, you really are focused on like the Llama 7B or Mixtral or Llama S-- uh, you know, 70B.

54. 10:11

    It's, it's kind of lo-- much lower quality models than what's available in the closed source world. Uh, so one of the things you can do is you can, um, implement what Google has called context caching.

55. 10:21

    In the, in the open source world, we'll, we'll have super long context models soon enough. But, uh, economically, right, you know, we talked about fifteen dollars token-- per million tokens output, um, and five million per to-- million tokens input.

56. 10:34

    If you were to have, uh, on, on, you know, the best o-- closed source models today, if you were to submit a prompt of like, you know, a million tokens, and, and most, most of the times you're looking at a document, you get a query back, right?

57. 10:45

    You- your, your output is very small. Almost all of the cost is just se-sending them that document, right? So that's, that's gonna really, really hurt you. So for people, you know, targeting maybe like a legal AI or like, um, you know, some sort of other contract review AI, a lot of these enterprise use cases, uh, prefill is

58. 11:02

    gonna dominate your cost if you're using APIs. Um, and so Google has this context caching and, and open source will have it so models you can run yourself and, and others will deploy over time.

59. 11:12

    Uh, but basically you don't recompute the KV cache, right, the, the context length every single time. Instead, you cache it. Uh, but the problem is to save, save that takes an, an incredible amount of memory.

60. 11:25

    Um, so you don't save it in the g-GPU's memory, right? You save it on the CPU's memory or storage. Um, and so, uh, vLLM, uh, which is an open source library for inference, is contributing-- is building this currently.

61. 11:39

    So if you're interested in contributing to that, uh, check that out. Um, or if you're interested in using it, just start the project, right? Um, because, you know- Well, most of the models we have in the closed source today are like only like 32 or 8K or 4K context length.

62. 11:52

    They're coming with longer, um, and being able to, you know, dramatically reduce your costs, um, by caching the context, um, is, is very-- is gonna, is gonna dramatically reduce cost, right?

63. 12:06

    Um, so now I'm just gonna talk about like head-in-the-cloud stuff instead of like real usable things, which is, um, you know, what's coming down the pipeline, right? Which is, you know, GPT-4 was like twenty thousand chips for ninety to a hundred days, um, used, you know, thirty-eight gigawatt hours.

64. 12:20

    Very, very expensive. Cool. Um, but, you know, what's-- what is, what are they building now, right? Uh, OpenAI, xAI, um, Anthropic, many others are building hundred thousand chip clusters, right?

65. 12:31

    And it would train GPT-4 in three days, right? So it's kind of irrelevant. Um, you know, and, and, uh, I'll, I'll skip over this part, uh, because it's not really, uh, too relevant.

66. 12:41

    Um, but, you know, what, what, what's a modern system capable of, right? Like H100 is, is pretty, uh, pretty fast relative to A100 and, and coming down the pipeline is these...

67. 12:50

    the new NVIDIA chips. But what, what, what's come-- you know, what's coming down with these hundred thousand GPU clusters, right? Um, it's not gonna be a one point eight trillion parameter model.

68. 12:58

    It's actually gonna be, you know, it could be in the tens of trillions of parameters. Um, you know, the, the training FLOPS, right? I talked about GPT-4 is it's roughly two e twenty-five FLOPS, right?

69. 13:08

    Which is, uh, you know, a number that's not really relevant or two e twenty-five FLOP. Um, but with a hundred thousand GPU cluster, you can do ten e twenty-six, ten e twenty-seven FLOPS.

70. 13:19

    Uh, and to run that model is gonna require two hundred gigabytes or terabytes a second of memory bandwidth, right? Um, but what does, what does that like-- what does that look like, right?

71. 13:26

    So, so this is a... On the top right is an image of, uh, M-Microsoft's data centers in Arizona where they're making GPT-5, right? Um, they have about a hundred thousand GPUs here.

72. 13:39

    Uh, it's a hundred and fifty megawatts, right? Like the average home does not consume... You know, that's like, that's like, like tens of thousands if not hundreds of thousands of homes of power consumption, right?

73. 13:49

    It's, it's kind of insane. Um, Elon's talked about his next generation cluster. He's building a hundred thousand GPU cluster today, uh, but he's talked about his next generation cluster is three hundred thousand GPUs.

74. 13:58

    That is kind of insane. But the, the power cost for that alone would be like five hundred million dollars a year, right? So it's like, you know, people are, people are kind of insane, but it's pretty cool.

75. 14:08

    Um, but, you know, the, the, the interesting thing here is, you know, on training, we, we, you know, when, when you, when you try and train a model today, people just talk about fully connected clusters.

76. 14:19

    Uh, every GPU is connected to every other GPU at some speed and you, you know, you have to do, you know, all your operations. But that's not really possible when you go to these super large clusters, right?

77. 14:29

    Um, so the hundred thousand GPU clusters, those are being built this year, and then next year they're planning to build multiple hundred thousand GPU clusters. Already you can see that it exists across multiple buildings, right?

78. 14:40

    Um, and so there's a lot of complicated networking, uh, going on, right, to connect these data centers together. Um, and, and one other thing I, that, that I think is just like kind of interesting to, again, head in the clouds just to think about is, um, when you connect these chips together, there's a lot of optics, right?

79. 14:57

    Uh, you know, you convert from electrical to optical, uh, and then, you know, over fiber optics to connect between chips, transceivers, et cetera, right? Uh, these are extremely unreliable, right?

80. 15:07

    Uh, they tend to have a failure rate around five years. Um, and so what's interesting is if you're talking about a hundred thousand GPU cluster, um, or if you're talking about a five hundred thousand GPU cluster, you're gonna have something fail like every five minutes, right?

81. 15:22

    Um, which is insane, right? How, how do you even deal with something in your cluster failing every five minutes when you're training a model, right? Um, so, you know, this is, this is again more of like a hardware-oriented thing but, uh, you know, the, the other thing that's interesting is like when you get chips, they're not all

82. 15:38

    the same speed. You know, an H100 is not an H100. Um, they're stragglers. Uh, so if you get a large distribution of chips, um, what we call it in the industry is, is called the silicon lottery, um, in that like, you know, you, you can buy, for example, a, a gaming GPU and, and compare it to other

83. 15:55

    people's gaming GPUs on the forums, and they're actually like percentages difference in performance. But when you do a massive training cluster, um, you end up with... You know, training is a synchronous workload, right?

84. 16:06

    You know, you, you, you update the weights, you then you pass the gradients around, right? Um, and then you, you know, then you again run through a bunch of data, uh, update the weights or pass the gradients around, update the weights, right?

85. 16:18

    Um, so it's, it's a synchronous workload. So if one of them is ten percent slower, then everything is ten percent slower. And ByteDance had a cool paper where actually they saw a twenty-five percent decrease in speed just because one random GPU they got, uh, while it did technically work, um, and NVIDIA...

86. 16:33

    and, and, and according to NVIDIA it was fine, it was like twenty-five percent slower than, uh, what they wanted, right? So they're... You know, this is like, this is on like a twenty thousand GPU cluster even, right?

87. 16:43

    Um, so, so it's, uh, it's, it's quite interesting that, you know, that, that's... These are the problems people are running into at scale, right? So they pulled that GPU out, um, and then you, you can sort of see their performance dramatically uplifted, right, um, during, during training.

88. 16:59

    Um, and then again, this is ByteDance on a twenty thousand GPU cluster. So it's, it's, um, it's a, it's a big, big issue. Um, and I think, I think some of the other stuff in this presentation is not really relevant.

89. 17:11

    Uh, but I think, I think what do these next generation systems look like is a very, um, important question to ask yourself, right? Um, you know, and what, what do I, what do I, what do I do when I deal with that, right?

90. 17:24

    Like I think a lot of the scaffolding that people are building, uh, today for LLMs are dealing with, you know, is, is dealing with hallucinations and things like that.

91. 17:33

    And, and the hope that everyone has, or at least a lot of the AGI people have is that, you know, when I, when I hundred X the compute, um, you know, when I build a cluster that takes five hundred million dollars of electricity and I train a model with it, it's gonna make something that, uh, uh, you

92. 17:47

    know, yearly ele-electricity costs and make a model with it. And then the cluster itself costs over ten billion by the way, right? Uh, it's, it's gonna get rid of a lot of these, um, hallucinations.

93. 17:56

    It's gonna let us do a lot of interesting things. Um, yeah. So, so I think that's, that's basically all for the talk. I just wanted to, you know, uh, mention, you know, sort of a reasonable thing which is how do you run Llama 405B kind of some strategies that people need to implement that aren't necessarily implemented yet,

94. 18:12

    uh, in the open source that are implemented at the labs. Um, but then also like, you know, what are they doing, right? 'Cause they're not worried about, you know, Llama 405B capable models. [outro music]