This is a fast, minimal implementation of Boris Dayma's DALL·E Mega. It has been stripped down for inference and converted to PyTorch. The only third party dependencies are numpy, requests, pillow and torch.
To generate a 4x4 grid of DALL·E Mega images it takes:
- 89 sec with a T4 in Colab
- 48 sec with a P100 in Colab
- 14 sec with an A100 on Replicate
- TBD with an H100 (@NVIDIA?)
The flax model and code for converting it to torch can be found here.
$ pip install min-dalleLoad the model parameters once and reuse the model to generate multiple images.
from min_dalle import MinDalle
model = MinDalle(is_mega=True, models_root='./pretrained')The required models will be downloaded to models_root if they are not already there. Once everything has finished initializing, call generate_image with some text and a seed as many times as you want.
text = 'Dali painting of WALL·E'
image = model.generate_image(text, seed=0, grid_size=4)
display(image)
text = 'Rusty Iron Man suit found abandoned in the woods being reclaimed by nature'
image = model.generate_image(text, seed=0, grid_size=3)
display(image)
text = 'court sketch of godzilla on trial'
image = model.generate_image(text, seed=6, grid_size=3)
display(image)
text = 'a funeral at Whole Foods'
image = model.generate_image(text, seed=10, grid_size=3)
display(image)
text = 'Jesus turning water into wine on Americas Got Talent'
image = model.generate_image(text, seed=2, grid_size=3)
display(image)
text = 'cctv footage of Yoda robbing a liquor store'
image = model.generate_image(text, seed=0, grid_size=3)
display(image)
Use image_from_text.py to generate images from the command line.
$ python image_from_text.py --text='artificial intelligence' --no-mega --seed=7
$ python image_from_text.py --text='trail cam footage of gollum eating watermelon' --mega --seed=1 --grid-size=3