For those who havent heard of pixart before, its claim to fame is being the first major model that is primarily LLM based input (It normally uses T5). So you can make really complex prompts with it, that you cant do with sdxl.

(And its also super small)

This model upload is just a straight conversion of https://huggingface.co/PixArt-alpha/PixArt-Sigma-XL-2-1024-MS to safetensors format.

This is mostly for StableUI's sake. There are still other steps you need to do for that, at

https://github.com/Stability-AI/StableSwarmUI/blob/master/docs/Model%20Support.md#pixart-sigma

There are additional steps you need to follow there.. but at least you can now skip the "convert from pth format to safetensors format" step.
Also, this way, you can use the built-in "download model from url" tool, because it only allows downloads of single files that are called ".safetensors"

The primary reason I'm uploading the base here, is that I also intend to release a finetune of it, and its kinda silly to have a finetune where the base is not equally available too.

License

Remember that this model is subject to the Pixart license.
(research only, no commercial use, I think?)