Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

cerebras.net

Post reply on HN
Print view View original

181–190 of 231 posts

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#181

Post by selcuka »

bilsbie wrote: post #111
2bitencryption wrote: post #105

Earlier quoted context omitted.

For the curious, here was the conversation I had: ME: What's another saying similar to: "The cat is out of the bag" or "The genie is out of the bottle"? chatgpt: Another similar saying is "the toothpaste is out of the tube." Like the other two expressions, this phrase conveys the idea that something has been revealed or unleashed, and it cannot be undone or put back into its original state. ME: Can you invent a new p…

I don’t understand why people aren’t more impressed with it clearly understanding and then even inventing idioms. That shows some real intelligence.

I'm definitely impressed when I see things like this. This is much more impressive than writing a 5 paragraph nonsense about a 2 sentence prompt.

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#183

Post by colordrops »

LightMachine wrote: post #164
wing-_-nuts wrote: post #161

Earlier quoted context omitted.

>So every time you're amazed by something chat-gpt4 says, remember that soon this will be in your pocket. I want to believe you, but I'm ignorant of the hardware requirements for these things. How soon do you think we'd be able to run something reasonably gpt4-like on, say, a 4090?

I feel like no less than 10 years if the singularity doesn't kick in before that. Hardware and energy isn't progressing as fast as we'd like, and that is the main bottleneck. As in, imagine a world where we actually had the same computing power required to train (not run) GPT-4 in 1s in a phone? That kind of world is way beyond AGI and the cure of cancer IMO. Which is great, because it gives us a very objective goal…

Your comment makes me wonder if it's not a coincidence that as we seem to be the hitting a limit in hardware power that human level intelligence begins to emerge.

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#184

Post by smaddox »

binarymax wrote: post #38

Here are the zero-shot accuracy numbers posted in the Huggingface evaluations for Cerebras-GPT 13B vs. the results of LLaMa 13B in their paper: Model BoolQ PIQA SIQA HellaSwag WinoGrande ARC-e ARC-c OBQA LLaMa 13B 78.1 80.1 50.4 79.2 73 74.8 52.7 56.4 Cerebras-GPT 13B - 76.6 - 51.3 64.6 71.4 36.7 28.6

Cerebras is "training compute optimal". Llama appears to be trained far beyond "training compute optimal". The tradeoff is that inference is closer to optimal for Llama, i.e. better performance with a smaller model.

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#185

Post by shagie »

tlb wrote: post #148
2bitencryption wrote: post #105

Earlier quoted context omitted.

For the curious, here was the conversation I had: ME: What's another saying similar to: "The cat is out of the bag" or "The genie is out of the bottle"? chatgpt: Another similar saying is "the toothpaste is out of the tube." Like the other two expressions, this phrase conveys the idea that something has been revealed or unleashed, and it cannot be undone or put back into its original state. ME: Can you invent a new p…

Conspiracy theory: this idiom used to be widespread on the internet, but a shadowy cabal of AI maximalists erased millions of such expressions from search engines so GPT could appear to re-invent them. It is a good expression though -- evocative but not gross or violent. You could imagine many less successful analogies to something ejecting something else.

That is not a conspiracy theory... it is alternate reality fan fiction.

While making up "what ifs" can be fun, it doesn't merit either of the words "conspiracy" or "theory".

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#186

Post by shagie »

brucethemoose2 wrote: post #35
arbuge wrote: post #24

Earlier quoted context omitted.

https://www.cerebras.net/product-chip/ There's a comparison picture there of one of their chips alongside a regular GPU chip. Effectively they use up the entire wafer.

Yeah, and that doesn't even do the nutty IO on these things justice. A 16x CS2 cluster like they describe is like a huge Nvidia cluster in terms of throughput, but more like a single Nvidia node structurally.

Don't forget the power consumption on them.

Cerebras Second-Gen Wafer Scale Chip: 2.6 Trillion 7nm Transistors, 850,000 Cores, 15kW of Power - https://www.tomshardware.com/news/cerebras-wafer-scale-engin...

Trying to cool 15kW to 20kW of power is also rather impressive. https://www.cerebras.net/cs2virtualtour - the engine block and cooling manifold

> The challenge of extracting more than 20 kW of heat from the wafer was solved by having the wafer "float" on a cold plate. The wafer is allowed to expand and contract while remaining in contact with the polished front side of the cold plate, despite the different coefficients of thermal expansion of copper and silicon. The cold plate is much more than a a slab of metal: advanced computational fluid dynamics modelling was used to design a labyrinth of coolant channels capable of maintaining a precise, stable, thermal environment even as 850,000 Al-optimized cores swing into action.

> The power density of the CS-2 is too high for direct air cooling, so liquid cooling is used instead. The internal manifold transfers heat between the CS-2 system's internal coolant and facilties water. Separating these two fluids ensure that the CS-2 system is not affected by changes in the quality of facilities water and that the very highest-quality coolant circulates through the cold plate.

> The two pump modules plug into the upper four dry-break connectors. The lower two are for the air-cooling or water-cooling heat exchanger.

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#187

Post by ShamelessC »

vintermann wrote: post #52
bogwog wrote: post #47

Earlier quoted context omitted.

In other words, they’re actually incentivized to help make LLMs as accessible as possible, rather than try to keep them locked up to hide them from competitors. Which makes me wonder if Nvidia is doing anything with LLMs too?

NVidia has certainly pushing the envelope on image generation. StyleGAN3 was really cool when it came. But it is an issue that their chips are hardly optimized for LLMs.

Have you seen the license on StyleGAN3?

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#188

Post by adt »

AlexanderTheGr8 wrote: post #182

Is there a regularly updated repository containing all the releases of LLMs as they happen? TBH I am tired of having to doommark (doom-bookmark) so many repositories and links...Would appreciate some collected database.

This is close, table of LLMs as released, and I try and add repos for the 'open' models: https://lifearchitect.ai/models-table/

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#189

Post by leereeves »

selcuka wrote: post #181
bilsbie wrote: post #111

Earlier quoted context omitted.

I don’t understand why people aren’t more impressed with it clearly understanding and then even inventing idioms. That shows some real intelligence.

I'm definitely impressed when I see things like this. This is much more impressive than writing a 5 paragraph nonsense about a 2 sentence prompt.

Is it?

There are many mentions of confetti cannons on the web, along with explanations of how they work (saying something like confetti shoots out of the cannon). Chat-GPT just picked a random thing (confetti) and completed the pattern "X out of Y" with the thing confetti comes out of.

It's easy. The cereal is out of the box. The helium is out of the balloon. The snow is out of the globe.

And it's exactly the one thing LLMs are trained to do: reproduce patterns of words. They are (perhaps) already better than humans at that one specific skill (another win for AI research) but I don't think it's a sign of general intelligence.

Re: Cerebras-GPT: A Family of Open, Compute-Efficient, Large Language Models

#190

Post by grishka »

cuuupid wrote: post #6
tombert wrote: post #2

Has anyone tried this? I have 96GB of GPU memory; will that be enough to run the biggest model?

13B fits nicely even in a 3090 (24gb vram)!

Does that mean I would be able to run it on an M1 Max with 64 gb unified memory?
Post reply on HN