search

lucidrains / deep-daze

Simple command line tool for text to image generation using OpenAI's CLIP and Siren (Implicit neural representation network). Technique was originally created by https://twitter.com/advadnoun
MIT License
4.37k stars 327 forks source link

Can We Put a VRAM warning at the top of the README? #86

Closed afiaka87 closed 3 years ago

afiaka87 commented 3 years ago

I have no idea where all this new traffic is coming from, but we've had like 4 posts in the past few days about trying to run this on consumer graphics cards. I think putting a note at or near the top of the README would be good at helping some of these folks. As it stands, the GPU optimizations are pretty far down a wall of text, so I can't exactly blame them for not getting to it. Maybe just something like:

FAQ: Q: I keep trying to run this and it's giving me an error about not having enough memory. What gives? A: This program requires a lot of VRAM. If you're trying to run it with less than 16 GB of VRAM on your card, please scroll down to see some options you can tinker with. In general, setting --image_width=256 will probably fix your problem if you have ~6-8 GB of VRAM. If you have less, then it's probably going to be impossible to run this program unfortunately.

afiaka87 commented 3 years ago

You could put a similar warning on big-sleep. As it stands, that program just barely blows 8 GiB I think. So a lot of people will be able to run it actually but they'll have to settle for an --image_size=256.

jack7803m commented 3 years ago

Not that it matters very much to the issue itself, but if you are curious, I believe a lot of this traffic is coming from TikTok. I've seen several people post about the program (each video with 30k+ likes), and they all just leave out how intensive it is to run, so adding a warning right at the top would be great.

afiaka87 commented 3 years ago

Not that it matters very much to the issue itself, but if you are curious, I believe a lot of this traffic is coming from TikTok. I've seen several people post about the program (each video with 30k+ likes), and they all just leave out how intensive it is to run, so adding a warning right at the top would be great.

Ah, that explains it. Thanks, I actually was kinda curious.

RaymondTerry commented 3 years ago

Hello, I am currently a university student looking for projects to contribute to. This project is very fascinating to me and the warning would also apply to my current setup as well seeing as I only have a GTX 1650 SUPER. Would it be okay if I make the the warning for the VRAM usage more apparent in the README file?

afiaka87 commented 3 years ago

That will be up to @lucidrains who I'm afraid is very busy currently with other projects. Thanks for the contribution though - we'll hopefully get around to merging it soon.