big-sleep VS DALLE2-pytorch

I am currently pursuing my Master's degree in Computer Science and I am interested in working on a deep learning model that can generate images based on text descriptions. I've been interested in the field for a long time (think google deep dream, or a few years back I was very into big-sleep)

Big Sleep

Other AIs that I don't have experience with but have heard good things about are DALL-E 2 and the open source Big Sleep.

Text-to-image, e.g. with Big Sleep

At first I didn't like the reason he claimed LaMDA was conscious because it seemed mostly based on the text output of the models, but listening to that made me realize he watered it down for the mainstream medias. And I actually had my own encounter with the consciousness of Big Sleep after using it a lot one day : it looked like a mass of eyes vaguely shaped like a rabbit, and kept showing me random images. I wouldn't be as convinced they have a consciousness if I didn't see it with my 3rd eye. But then again, I also found evidence that regular non AI programs can develop one too so who know.

and after a few hours got this: https://i.imgur.com/FxdfdmV.png

Not nearly as cool as the real DALL-e, but maybe I'm missing something.

[1] https://github.com/lucidrains/big-sleep

(Added Mar. 23, 2021) Big Sleep - Colaboratory by LtqxWYEG. Uses BigGAN to generate images. Reference.

I was showing people Dalle2 last year and telling them how much of an impact an open source solution was going to have on, well, everything to do with art and design. (At the time Stable Diffusion had not released, not even the leak, and all hopes was on https://github.com/lucidrains/DALLE2-pytorch)

Stable diffusion is the model, MJ will use a variant and DALL-E is the old version (we have our own implementation from our distinguished fellow Lucidrains here: https://github.com/lucidrains/DALLE2-pytorch)

Sadly I don't think so. From what I understand the architecture is fixed to 1024x1024 pictures.

The code is open source, the model is not I believe. https://github.com/lucidrains/DALLE2-pytorch

Phil Wang has been fairly prolific at creating open source implementations of these text to image models. For example, here is the dalle-2 repo https://github.com/lucidrains/DALLE2-pytorch

There's already an open-source implementation of DALL-E 2 (https://github.com/lucidrains/DALLE2-pytorch) and a pretrained model for it should be released within this year.

Also true for Google's Imagen, which should be even better than DALLE-2 (and faster) https://github.com/lucidrains/imagen-pytorch.

This is possible because the original research papers behind both DALLE-2 and Imagen were publicly released.

The paper describing the model is public and has been implemented here, but that's not the hard part. The model likely requires months of compute and dozens of gigabytes of VRAM to train and run, likely costing several hundred thousand dollars.