dalle-mini

Nice pictures. Really wish I'd started my GenAI benchmark back when DALL-E 2 were still around, just to have them as a fun little baseline for how much things have changed. The only DALL-E 2 images [1] I could dig up in my archives that matched any of the bench prompts was the Alexander the Great prompt: A painting of Alexander the Great riding a hippity hop into battle. The hippity hop toy was an old children’s toy that looked like a giant rubber ball with rubber handles that one straddle and ride. And while it's not exactly the same, DALL-E Mini [2] (from back in 2021) is still around and you can run it locally using pretty modest hardware. [1] - https://imgpb.com/ZyzuX [2] - https://github.com/borisdayma/dalle-mini

Mini-Gemini is a bit of a confusing name. Reminds me of how DALL·E Mini came out three years ago and eventually had to rename itself to Craiyon https://github.com/borisdayma/dalle-mini

Co-created Dall-E Mini (now named Craiyon, and I am not involved anymore). It was among the first text to language models created independently. And it was fully open source. It also got covered by New York Times in the article covering Dall-E 2 by Cade Metz. Links: - GitHub: https://github.com/borisdayma/dalle-mini - Hugging Face Demo: https://huggingface.co/spaces/flax-community/dalle-mini - NYT article: https://www.nytimes.com/2022/04/06/technology/openai-images-... ___ (I know this is not as much impactful as others in this thread. But I did this after less than 2 years after transitioning to tech from Physics, and at the age of 22.)

You can run craiyon / dalle-mini on a card with 8GB of VRAM if you decrease batch size to 1 and skip the CLIP step. Takes about 7 sec to generate an image on a 3070. I started with https://github.com/borisdayma/dalle-mini/blob/main/tools/inf... and pared it down.

Wow, this author is very dishonest as it does not mention any of the people who created this project in the first place. I was one of the people who worked in this project. This was spearheaded by Boris Dayma, now at Weights and Biases. This is an Open Source project with all code and methods in public. See either GitHub (https://github.com/borisdayma/dalle-mini) or the hosted space in Hugging Face Hub (https://huggingface.co/spaces/dalle-mini/dalle-mini) or the project report (https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-G...). This project was also covered in the NYT article on Dalle2 by Cade Metz. The author gives no credits at all. That is apalling. (Also, the one hosted in the HF Hub gives you better results) I just realized that this person is either using our model (some point in the past) and not giving us due credit

https://github.com/borisdayma/dalle-mini https://rudalle.ru/en/ Smaller reproductions of the original research.