Model or just t5xxl?
Is this the flux1 merged with t5xxl ? Or I'd it just the t5xxl model? if it is can you please quantize flux 1 base model for us that have limited resources we need quantize 8,5,and 4 base model
These are only t5xxl.
I have not tested these quantized models to work, so I'll upload the quantized transformer block after I check them.
You can try quantizing with optimum-quanto if you want to do it yourself.
https://github.com/huggingface/optimum-quanto?tab=readme-ov-file#diffusers-models
Can I do this on cpu in termux android or do I need PC? Nvidia cuda gpu?
NVIDIA GPU card is not required, but a desktop PC with large RAM is preferred. I think it is maybe hard to run on Android.
Can you please make a YouTube video showing how to use optimo-qauntinization?
Idk if your able to record while qauntinizing maybe you can try screen shots
The quantizing code is just same as on the README. If you don't have any coding skills, it's good to wait for other community people to quantize them.
Some people are already trying quantizing transformer block, so maybe it won't take so long to appear.
https://huggingface.co/lllyasviel/flux1-dev-bnb-nf4