electronsoup

26 giorni fadiscuss

Yeah MoE is a little worse for the same size, but you can often run bigger MoEs at respectable speeds even on cpu ram offload. The dense models really need to be 100% vram

electronsoup

26 giorni fadiscuss

> It gets into loops quite often, and surprisingly often gets the edit tool call wrong

I find that running better quantization, like Q8 tend to prevent this even though its a bit slower to run, it saves overall time with less churn

Using 3.6-27b is even slower again than 3.6-35b, but I find the accuracy really pays off

electronsoup

29 giorni fadiscuss

> in secret is impossible without the whole world knowing.

I'm curious about why this is

Outside of an actual test detonation, presumably this could all happen in a secure place?

electronsoup

29 giorni fadiscuss

They didn't say they had never traveled south though

electronsoup

30 giorni fadiscuss

If you put that behind an API, you could sell the service much like the AI providers

electronsoup

mese scorsodiscuss

Why this is not a PR for llama.cpp

electronsoup

2 mesi fadiscuss

> likely not well thought out

Or it has been, and cruelty is the point

electronsoup

2 mesi fadiscuss

Surely this will get arbitraged like anything else, where fans who get picks will onsell tickets

electronsoup

2 mesi fadiscuss

So now we need to run farms of spotify accounts playing songs to get our concert tickets?

electronsoup

2 mesi fadiscuss

You may need to move on to other services like Apple Music

electronsoup

2 mesi fadiscuss

So how many of their employees are now familiar with the codebase? zero?

electronsoup

3 mesi fadiscuss

At what point does spanish internet become too unreliable? There was a thread the other day about someone's CI jobs failing due too this.

electronsoup

3 mesi fadiscuss

> the GPU apps we are building with them are

I can't help but get the feeling you have use-case end-goal in mind that's opaque to many of us who are gpu-ignorant.

It could be helpful if there were an example of the type of application that would be nicer to express through your abstractions.

(I think what you've shown so far is super cool btw)

electronsoup

3 mesi fadiscuss

> worth their salt

That's a big assumption. Often there's no time to do things right, or no money, or lack of oversight, and so on.

Not every company is staffed by empowered and highly motivated staff

electronsoup

3 mesi fadiscuss

and most OS do enable it by default

electronsoup

3 mesi fadiscuss

Whenever I see a chip like this, I think "why wont my company let me use a decent computer"

electronsoup

4 mesi fadiscuss

Which mainboards are cheap and have 4 pcie16x (electrical) slots, that don't need weird risers to fit 4 GPUs

electronsoup

4 mesi fadiscuss

If it was so important, wouldn't he just filibuster it till he got what he wanted?

electronsoup

4 mesi fadiscuss

> but then again you need to plausibly explain why was someone operating your car while you were not aware of it.

There is no such requirement.

electronsoup

4 mesi fadiscuss

> and they're just as capable as using AI as anyone

Wouldn't the assumption be the opposite, in that AI is magnifying the decision making of the engineer and so you get more payback by having the senior drive the AI?

electronsoup

Submissions

BugStalker for Rust

comments