194

The first GPT-4-class AI model anyone can download has arrived: Llama 405B(arstechnica.com)

posted
by
Avatar
Wilshire@lemmy.world
in
Avatar
technology@lemmy.world
61 comments
hidereport
Sort:
HotTopControversialNewOld
You are viewing a single thread.
View all comments
Avatar
just another dev@lemmy.my-box.dev
97 points

Technically correct ™

Before you get your hopes up: Anyone can download it, but very few will be able to actually run it.

permalink
report
reply
Avatar
bitfucker@programming.dev
5 points

So does OSM data. Everyone can download the whole earth but to serve it and provide routing/path planning at scale takes a whole other skill and resources. It’s a good thing that they are willing to open source their model in the first place.

permalink
report
parent
reply
Avatar
LavenderDay3544@lemmy.world
8 points
*

When the RTX 9090 Ti comes, anyone who can afford it will be able to run it.

permalink
report
parent
reply
Avatar
Contravariant@lemmy.world
3 points

That doesn’t sound like much of a change from the situation right now.

permalink
report
parent
reply
Avatar
coffee_with_cream@sh.itjust.works
12 points
*

This would probably run on a a6000 right?

Edit: nope I think I’m off by an order of magnitude

permalink
report
parent
reply
Avatar
5redie8@sh.itjust.works
2 points

“an order of magnitude” still feels like an understatement LOL

My 35b models come out at like Morse code speed on my 7800XT, but at least it does work?

permalink
report
parent
reply
Avatar
chiisana@lemmy.chiisana.net
23 points

What’s the resources requirements for the 405B model? I did some digging but couldn’t find any documentation during my cursory search.

permalink
report
parent
reply
Avatar
sunzu@kbin.run
12 points

405b ain’t running local unless you got a proepr set up is enterpise grade lol

I think 70b is possible but I haven’t find anyone confirming it yet

Also would like to know specs on whoever did it

permalink
report
parent
reply
Avatar
raldone01@lemmy.world
2 points
*

I regularly run llama3 70b unqantized on two P40s and CPU at like 7tokens/s. It’s usable but not very fast.

permalink
report
parent
reply
Show more comments
Avatar
Voyajer@lemmy.world
8 points
*

I’ve run quantized 70B models on CPU with 32 gigs but it is very slow

permalink
report
parent
reply
Show more comments
Avatar
modeler@lemmy.world
41 points
*

Typically you need about 1GB graphics RAM for each billion parameters (i.e. one byte per parameter). This is a 405B parameter model. Ouch.

Edit: you can try quantizing it. This reduces the amount of memory required per parameter to 4 bits, 2 bits or even 1 bit. As you reduce the size, the performance of the model can suffer. So in the extreme case you might be able to run this in under 64GB of graphics RAM.

permalink
report
parent
reply
Avatar
Longpork3@lemmy.nz
1 point
*

Hmm, I probably have that much distributed across my network… maybe I should look into some way of distributing it across multiple gpu.

Frak, just counted and I only have 270gb installed. Approx 40gb more if I install some of the deprecated cards in any spare pcie slots i can find.

permalink
report
parent
reply
Avatar
obbeel@lemmy.eco.br
2 points

According to huggingface, you can run a 34B model using 22.4GBs of RAM max. That’s a RTX 3090 Ti.

permalink
report
parent
reply
Avatar
arefx@lemmy.ml
1 point

Ypu mean my 4090 isn’t good enough 🤣😂

permalink
report
parent
reply
Avatar
cheddar@programming.dev
22 points

Typically you need about 1GB graphics RAM for each billion parameters (i.e. one byte per parameter). This is a 405B parameter model.

permalink
report
parent
reply
Avatar
Siegfried@lemmy.world
8 points
*

At work we habe a small cluster totalling around 4TB of RAM

It has 4 cooling units, a m3 of PSUs and it must take something like 30 m2 of space

permalink
report
parent
reply
Avatar
Deceptichum@quokk.au
13 points

Or you could run it via cpu and ram at a much slower rate.

permalink
report
parent
reply
Show more comments
Avatar
1984@lemmy.today
3 points

Can you run this in a distributed manner, like with kubernetes and lots of smaller machines?

permalink
report
parent
reply
Avatar
TipRing@lemmy.world
4 points

When the 8 bit quants hit, you could probably lease a 128GB system on runpod.

permalink
report
parent
reply
Avatar
Blaster M@lemmy.world
6 points

As a general rule of thumb, you need about 1 GB per 1B parameters, so you’re looking at about 405 GB for the full size of the model.

Quantization can compress it down to 1/2 or 1/4 that, but “makes it stupider” as a result.

permalink
report
parent
reply

Technology

!technology@lemmy.world

Create post

This is a most excellent place for technology news and articles.

Our Rules

  1. Follow the lemmy.world rules.
  2. Only tech related content.
  3. Be excellent to each another!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, to ask if your bot can be added please contact us.
  9. Check for duplicates before posting, duplicates may be removed

Approved Bots

Community stats

  • 17K

    Monthly active users

  • 6.1K

    Posts

  • 130K

    Comments

Community moderators