yeah it seemed comparable if i keep it at 75 tokens on the GPT3 tokenizer i never get the error
I’m running a 5800x and RTX 3080
I can render one sample at 768p in 18 seconds or at 512p in 6 seconds
yeah i’ve heard it doesn’t use the GPU on the M1, i didn’t even bother installing it on my Macbook lol.. I heard stability is working on Mac and AMD support tho
afaik people are using mps with better results. it requires pytorch nightly versions atm. think i saw someone say they generated a 512x image in 30 sec, but it was probably the top end m1.
10 minutes was still wild tho since it was literally just cpu. idk enough about the low level hardware stuff but my understanding is software optimizations alone in pytorch, tf, etc are having major benefits for apple silicon
afaik people are using mps with better results. it requires pytorch nightly versions atm. think i saw someone say they generated a 512x image in 30 sec, but it was probably the top end m1.
10 minutes was still wild tho since it was literally just cpu. idk enough about the low level hardware stuff but my understanding is software optimizations alone in pytorch, tf, etc are having major benefits for apple silicon
oh interesting i didn’t even know stablediffusion / pytorch supported mps
i have the M1 Max 24 core 32GB might be worth an install
yeah that makes sense yeah i don’t know much about low level hardware stuff either
oh interesting i didn’t even know stablediffusion / pytorch supported mps
i have the M1 Max 24 core 32GB might be worth an install
yeah that makes sense yeah i don’t know much about low level hardware stuff either
oh yeah, definitely look into it then.
I think Apple has been pushing for better support from tensor flow and PyTorch etc for a while idk
But the latest nightly versions of pytorch support mps. I think there’s an issue in the compvis sd gh repo with pretty simple instructions now
!dream Man riding a high-speed motorbike through a gloomy dark green tunnel at night, 4k quality, highly detailed, low angle photography

I made this with Dall-E 2 today, took me like 40 prompts..also used a lot the Outpainting mode.
Had to use Photoshop to add and combine some stuff (like the ghost, the swirly effect and some borders of the final product)
The prompt was:
Man riding a high-speed motorbike through a gloomy dark green tunnel at night, 4k quality, highly detailed, low angle photography.
Started with this:
End result:
!dream Man riding a high-speed motorbike through a gloomy dark green tunnel at night, 4k quality, highly detailed, low angle photography.

!dream Man riding a high-speed motorbike through a gloomy dark green tunnel at night, 4k quality, highly detailed, low angle photography

it goes without saying that dalle2 will generally give better results with less effort, but you can’t really compare prompts between models besides simple ones like an astronaut riding a horse
when I first tried de after months of midjourney I was comparing prompt for prompt and was really underwhelmed
!dream Man riding a high-speed motorbike through a gloomy dark green tunnel at night, 4K neon lights colors detailed, low angle photography, artstation

!dream stunning photo of a man riding a high-speed motorbike through a neon lit tunnel at night, 4K, lights colors vivid, detailed, low angle photography

someone do
!dream man eating a cheeseburger while riding a bike through a tokyo city alleyway during the afternoon, 4k neon lights colors detailed, low angle photography
!dream 16mm photograph, deep focus, brass sculpture of the creator of the universe, picasso

someone do
!dream man eating a cheeseburger while riding a bike through a tokyo city alleyway during the afternoon, 4k neon lights colors detailed, low angle photography
!dream man eating a cheeseburger while riding a bike through a tokyo city alleyway during the afternoon, 4k neon lights colors detailed, low angle photography

!dream man eating a cheeseburger while riding a bike through a tokyo city alleyway during the afternoon, 4k neon lights colors detailed, low angle photography

f***, shouldve specified with one hand eating cheesburger, with other holding the handlbars. he looks like a gorgon
f***, shouldve specified with one hand eating cheesburger, with other holding the handlbars. he looks like a gorgon
drop $4 bro đź‘€
f***, shouldve specified with one hand eating cheesburger, with other holding the handlbars. he looks like a gorgon
!dream man eating a cheeseburger with one hand while other holds handlebar, riding bike through a tokyo city alleyway, afternoon, 4k neon lights colors detailed, low angle photography

oh yeah, definitely look into it then.
I think Apple has been pushing for better support from tensor flow and PyTorch etc for a while idk
But the latest nightly versions of pytorch support mps. I think there’s an issue in the compvis sd gh repo with pretty simple instructions now
yeah I now see Apple added pytorch support for mps in June
nice had no idea I could do fast pytorch ML workloads
found the issue and steps, ty
yeah i will install it soon, i have my gaming pc running a local web server at the moment so i haven’t thought much about it but it would be nice to have a way to run it away from home
!dream tilt-shift photography, highly-detailed clay model of village in the middle of the jungle, green and blue tones, night, heavy rain, bright moonlight

yep. Sorry if that wasn’t clear
AYOOOO dropping $4 way more reasonable now