CUDA OUT OF MEMORY | Invoke | Page 1

high lake Dec 17, 2022, 4:56 PM

#

I have all requeriments

scarlet needle Dec 17, 2022, 5:00 PM

#

You're trying to run it on a 4GB card?

#

Is that what I'm reading there?

high lake Dec 17, 2022, 5:02 PM

#

I see others with 6gb card, and having the same poroblem

#

the real problem is the 4Gb card?

scarlet needle Dec 17, 2022, 5:03 PM

#

I would agree there, I think 8GB is the bare minimum

#

Probably. I've been here like 5 days, let me double check

high lake Dec 17, 2022, 5:03 PM

#

omg

scarlet needle Dec 17, 2022, 5:04 PM

#

I mean I have an 8G card and it's not enough to train a model, got me looking at Nvidia Tesla K80 24G cards wondering if they'll work lol

high lake Dec 17, 2022, 5:04 PM

#

omggg

#

T_T

scarlet needle Dec 17, 2022, 5:04 PM

#

3070 ti is a cheap way to get in on this though.

#

And it'll render in like 5 seconds

#

Or use a cloud-based solution

high lake Dec 17, 2022, 5:05 PM

#

google colab?

scarlet needle Dec 17, 2022, 5:05 PM

#

I'll leave those suggestions to the pros.

#

I just really like it here so I try to give back a little and answer stuff

high lake Dec 17, 2022, 5:06 PM

#

i understand

scarlet needle Dec 17, 2022, 5:07 PM

#

I'm researching how the Nvidia Tesla cards work. I might buy a couple and see if they actually work for this stuff. They're like $100 new / $60 questionably used.

#

And 24G (2x12GB)

#

But like I said I've been here at most a week

high lake Dec 17, 2022, 5:08 PM

#

yeah these models are not that expensive

#

but anyway, thanks for your attention and help

tawdry viper Dec 17, 2022, 6:24 PM

#

the bare minimum for InvokeAI is 4gb BUT i think that depends a lot upon if your card can support fp16

#CUDA OUT OF MEMORY