gaeld

9 ms·

1.
by gaeld 4mo ago
True, and for third-party models we'll just re-use their public open weights. There is a time-consuming part, though, that is performed manually by our (human) team: implement the logic of the model in C++ and assembly code in a super-
2.
by gaeld 4mo ago
It also matters for thinking models and for agentic workflows, especially in software engineering, where a lot of tokens need to be output in iterative loops before the user sees any result. This is our main use case.
3.
by gaeld 4mo ago
In theory yes, although not in a linearly proportional way, because in practice our memory streaming is not yet perfect. There are still some fixed costs that we did not fully optimize (for now).
4.
by gaeld 4mo ago
I'm sure there are, and I really hope we can work on consumer-grade GPUs at some point. It should be possible to apply the same methodology (digging deep into the hardware details to understand all its little characteristics, and rethi
5.
by gaeld 4mo ago
Note that this coding model is trained on programming use cases, and is also not tuned for multi-turn chat. You can ask it to implement an algorithm; we provide suggested prompts you can test. Also, this tech preview is really about the spe
6.
by gaeld 4mo ago
Why not, it's one way to look at it! Although I have yet to see other work with speculative decoding higher than ~1,000 tokens/s., because the other bottlenecks start to matter at that point, and they need to be solved to go furth
7.
by gaeld 4mo ago
it's also a coding model
8.
by gaeld 4mo ago
thank you deflator, I understand this now! much appreciated
9.
by gaeld 4mo ago
Token generation speed matters for sequential agentic workflows, like software engineering / vibe coding, where a lot of reasoning tokens, code generation, refactoring, testing, etc. happen in a loop before an actual outcome is served
10.
by gaeld 4mo ago
will do - we are a small team and it takes time to implement and optimize a new model, whatever the size.
11.
by gaeld 4mo ago
Totally, though DTP is not required for these kind of speeds. Standard TP works also. DTP is something we built for our roadmap in order to get to extremely high speeds (like 10k+ tokens/s). When the budget is under 10 µs per layer, an
12.
by gaeld 4mo ago
Thanks for the comment and the question! The last section of the article lays out the scaling laws that apply when porting this approach to another model. In a nutshell, DeepSeek V4 Pro with 49B active params is close to the upper bound. Al
13.
by gaeld 4mo ago
Thanks a lot! Much appreciated. To answer your questions: - yes, we rewrite the whole model code (while keeping the same logic) in CUDA/HIP and assembly, in order to optimize by hand for each GPU type. It's quite tedious for sure,
14.
by gaeld 4mo ago
Yeah, I agree: I'm actually not expecting it to be easy, and there will certainly be several unknown unknowns we'll discover along the way. Our process has been, and will continue to be, a sequence of (tedious) R&D experiments
15.
by gaeld 4mo ago
Great points, let me clarify: - model size: 2B is just for this preview (it was faster to implement), our article explains how we expect to support large frontier MoE at 1,000 to 5,000 tokens/s - reaching 500 tok/s, or even up to
16.
by gaeld 4mo ago
I updated the article title accordingly
17.
by gaeld 4mo ago
YES - I just updated the title of our article according to your suggestion.
18.
by gaeld 4mo ago
Follow-up reading the most technical and research people here: Monokernel deep dive (GPU Engineering): http://blog.kog.ai/building-a-single-kernel-latency-optimize... Delayed Tensor Parallelism (research): http:/&#x2F
19.
by gaeld 4mo ago
I guessed you thought about consumer GPUs. We are about standard datacenter GPUs indeed. Sorry for the confusion
20.
by gaeld 4mo ago
I guessed you thought about consumer GPUs. We are about standard datacenter GPUs indeed.
21.
by gaeld 4mo ago
thanks! we explain how it scales to larger models in the last section the OP blog post
22.
by gaeld 4mo ago
Thanks. To be fair, this number is what we expect to get once we port DeepSeek V4 in our engine on the upcoming generation of GPUs!
23.
by gaeld 4mo ago
Great points. We strived to be fair as possible in the benchmark, but it's indeed not perfect. Taalas should have been added in the dedicated hardware section, even though they use 3-bit quantization when we are on FP16 (to be fair in
24.
by gaeld 4mo ago
Fair point - this tech preview is about the speed (hence the small dense model, it was easier to implement). The math checks out though to allow support for large frontier MoE models at similar speeds. At batch size 1, GPT-OSS-120B has 5.1B