search

Alpha-VLLM / Lumina-T2X

Lumina-T2X is a unified framework for Text to Any Modality Generation
MIT License
1.81k stars 73 forks source link

ComfyUI support? #68

Open GamingDaveUk opened 1 week ago

GamingDaveUk commented 1 week ago

Is there a way to use this in comfyui? Really impressed with the prompt following that a user shared into a discord channel.

Also can lora's be created for it? Can it be trained?

C0nsumption commented 1 week ago

Interested as well （づ￣3￣）づ

kijai commented 1 week ago

Made a wrapper that can run the T2I models: https://github.com/kijai/ComfyUI-LuminaWrapper

GamingDaveUk commented 1 week ago

Made a wrapper that can run the T2I models: https://github.com/kijai/ComfyUI-LuminaWrapper

Very cool, though I am on Windows and slightly reluctant to use pre-built wheels (the llmvision* issue has me cautious)

*think that was the name of it.

kijai commented 1 week ago

Made a wrapper that can run the T2I models: https://github.com/kijai/ComfyUI-LuminaWrapper

Very cool, though I am on Windows and slightly reluctant to use pre-built wheels (the llmvision* issue has me cautious)

*think that was the name of it.

I get that, but it's the same issue with the original code. There's a fallback to SDP attention in the code, but it does not work at all. I built flash_attn on my Windows install and it works fine, took almost an hour though...

Excidos commented 1 week ago

I'm having a lot of trouble with flash attention on my windows, any help would be greatly appreciated :)

kijai commented 1 week ago

I'm having a lot of trouble with flash attention on my windows, any help would be greatly appreciated :)

It now works without flash_attn, it's just much slower and uses twice as much VRAM, but it works.

PierrunoYT commented 1 week ago

I need a something to host a Demo Locally on my 4090