Zebra-Llama – Towards efficient hybrid models

by mirrir • 8 months ago

113 61 8 months ago

mason_mpls• 8 months ago

Zebra-Llama achieves Transformer-level accuracy with near-SSM efficiency using only 7–11B training tokens (compared to trillions of tokens required for pre-training) and an 8B teacher. Moreover, Zebra-Llama dramatically reduces KV cache size—down to 3.9%, 2%, and 2.73% of the original for the 1B, 3B, and 8B variants, respectively—while preserving 100%, 100%, and 97% of average zero-shot performance on LM Harness tasks.

This is an extraordinary claim, is there a catch I’m missing? Am I misreading?

jychang8• 8 months ago

The catch that you're missing is that Deepseek did this ages ago.

They're just using MLA, which is well known to reduce KV size by 90%. You know, the MLA that's used in... Deepseek V2, Deepseek V3, Deepseek R1, Deepseek V3.1, Deepseek V3.2.

Oh, and they also added some hybrid linear attention stuff to make it faster at long context. You know who else uses hybrid linear attention? Deepseek V3.2.

erichocean• 8 months ago

Kimi K2 also uses MLA, and Kimi Linear runs Kimi Delta Attention (it's SSM-like) for three out of every four layers (the fourth uses MLA).

storus• 8 months ago

Linear attention is really bad, it's only good for benchmaxing but it leads to a loss of valuable granularity, which can be felt in the latest DeepSeek randomly forgetting/ignoring/correcting explicitly stated facts in the prompt.

adityashankar3• 8 months ago

Due to perverse incentives and the historical nature of models over-claiming accuracy, it's very hard to believe anything until it is open source and can be tested out

that being said, I do very much believe that computational efficiency of models is going to go up [correction] drastically over the coming months, which does pose interesting questions over nvidia's throne

*previously miswrote and said computational efficiency will go down

danielbln• 8 months ago

I think you mean computational efficiency will go _up_ in the future. To your last point: Jevons paradox might apply.

adityashankar3• 8 months ago

yup that's what I meant!, Jevon's paradox applies to resource usage in general and not towards a specific companies dominance

if computational efficiency goes up (thanks for the correction), and CPU inference becomes viable for most practical applications, GPUs (or accelerators) themselves may be unnecessary for most practical functions

credit_guy4• 8 months ago

Like this?

https://huggingface.co/amd/Zebra-Llama-8B-8MLA-24Mamba-SFT

adityashankar3• 8 months ago

yes!, thanks for the link!

moffkalast• 8 months ago

GGUF when? /s

deepdarkforest• 8 months ago

which does pose interesting questions over nvidia's throne...
Zebra-Llama is a family of hybrid large language models (LLMs) proposed by AMD that...

Hmmm

jychang8• 8 months ago

Or like this: https://api-docs.deepseek.com/news/news251201

I don't know what's so special about this paper.

- They claim to use MLA to reduce KV cache by 90%. Yeah, Deepseek invented that for Deepseek V2 (and also V3 and Deepseek R1 etc)

- They claim to use a hybrid linear attention architecture. So does Deepseek V3.2 and that was weeks ago. Or Granite 4, if you want to go even further back. Or Kimi Linear. Or Qwen3-Next.

- They claimed to save a lot of money not doing a full pre-train run for millions of dollars. Well, so did Deepseek V3.2... Deepseek hasn't done a full $5.6mil full pretraining run since Deepseek V3 in 2024. Deepseek R1 is just a $294k post train on top of the expensive V3 pretrain run. Deepseek V3.2 is just a hybrid linear attention post-train run - i don't know the exact price, but it's probably just a few hundred thousand dollars as well.

Hell, GPT-5, o3, o4-mini, and gpt-4o are all post-trains on top of the same expensive pre-train run for gpt-4o in 2024. That's why they all have the same information cutoff date.

I don't really see anything new or interesting in this paper that isn't already something Deepseek V3.2 has already sort of done (just on a bigger scale). Not exactly the same, but is there anything amazingly new that's not in Deepseek V3.2?

ACCount372• 8 months ago

I don't doubt the increase in efficiency. I doubt the "drastically".

We already see models become more and more capable per weight and per unit of compute. I don't expect a state-change breakthrough. I expect: more of the same. A SOTA 30B model from 2026 is going to be ~30% better than one from 2025.

Now, expecting that to hurt Nvidia? Delusional.

No one is going to stop and say "oh wow, we got more inference efficiency - now we're going to use less compute". A lot of people are going to say "now we can use larger and more powerful models for the same price" or "with cheaper inference for the same quality, we can afford to use more inference".

colechristensen5• 8 months ago

Eh.

Right now, Claude is good enough. If LLM development hit a magical wall and never got any better, Claude is good enough to be terrifically useful and there's diminishing returns on how much good we get out of it being at $benchmark.

Saying we're satisfied with that... well how many years until efficiency gains from one side and consumer hardware from the other meet in the middle so "good enough for everybody" open models are available for anyone who wants to pay for a $4000 MacBook (and after another couple of years a $1000 MacBook, and several more and a fancy wristwatch).

Point being, unless we get to a point where we start developing "models" that deserve civil rights and citizenship, the years are numbered to where we NEED cloud infrastructure and datacenters full of racks and racks of $x0,000 hardware.

I strongly believe the top end of the S curve is nigh, and with it we're going to see these trillion dollar ambitions crumble. Everybody is going to want a big-ass GPU and a ton of RAM but that's going to quickly become boring because open models are going to exist that eat everybody's lunch and the trillion dollar companies trying to beat them with a premium product aren't going to stack up outside of niche cases and much more ordinary cloud compute motivations.

Reubend• 8 months ago

It would be REALLY cool to see this same technique applied to a much more recent OSS model distillation. For example, Mistral 3 14B would be a great target. How efficient can we get inference there?

a_wild_dandan• 8 months ago

If the claims in the abstract are true, then this is legitimately revolutionary. I don’t believe it. There are probably some major constraints/caveats that keep these results from generalizing. I’ll read through the paper carefully this time instead of a skim and come back with thoughts after I’ve digested it.

jychang8• 8 months ago

What's not to believe? Qwerky-32b has already done something similar as a finetune of QwQ-32b but not using traditional attention architecture.

And hybrid models aren't new, MLA based hybrid models is basically just Deepseek V3.2 in a nutshell. Note that Deepseek V3.2 (and V3.1, R1, and V3... and V2 actually) all use MLA. Deepseek V3.2 is what adds the linear attention stuff.

Actually, since Deepseek V3.1 and Deepseek V3.2 are just post-training on top of the original Deepseek V3 pretrain run, I'd say this paper is basically doing exactly what Deepseek V3.2 did in terms of efficiency.

cubefox• 8 months ago

DeepSeek-V3.2 is a sparse attention architecture, while Zebra-Llama is a hybrid attention/SSM architecture. The outcome might be similar in some ways (close to linear complexity) but I think they are otherwise quite different.

xer• 8 months ago

This is great! But what if the US invests 1% of GDP in GPU datacenters and then those are not needed becaues someone created a much more efficient architecture?

_boffin_• 8 months ago

They will still use capacity. Why would you believe anything different?

wild_egg• 8 months ago

More efficiency just means more consumption. Think when they add lanes to a highway, traffic gets better for a little bit but very soon the highway is just as congested as before.

wilg• 8 months ago

More people get where they’re going in the same amount of time though

dkural• 8 months ago

Look up Jevons Paradox, when something becomes more efficient, consumption can goes up, often due to price elasticity.

Think of like this: Imagine car prices go from $200,000 to $$20,000 - you wouldn't sell 10x the amount of cars, you'd sell --- In fact I just looked up the numbers - worldwide only 100K or so cars are 200K & higher, whereas roughly 80 million cars are in that affordable category.

So a price drop of 90% allowed sales to go from 0.1M to 80M!! I think this means we need more engines, tires, roads, gas, spare parts.

chpatrick• 8 months ago

Then they'll be able to use those datacenters much more efficiently.

AlexCoventry• 8 months ago

This is from May 2025, according to the arxiv watermark. Maybe that should be mentioned in the title.

KnuthIsGod• 8 months ago

Looks like the trillions of dollars spent on datacentres will end up being regretted.

pryelluw• 8 months ago

I should have been an electrician.

Zebra-Llama – Towards efficient hybrid models

Related Stories