All Zeros/Black Image
I am using the official Diffusers script to load the model offline, but I get a completely black image. Has anyone encountered the same issue? How to solve it?
do you use sage-attention?
do you use sage-attention?
thanks. I solved it
I have same issue. I do not use sage-attention, it's not even installed on my system. I'm using the sample diffusers code as a starting point with an RTX 3060 12GB and call enable_sequential_cpu_offload() after loading the pipeline. other than that, it's basically the same code as the example code on your model card. This is on Debian bookworm, python 3.13 and the latest diffusers in a fresh conda environment.
I encountered the same situation, how did you solve it？
I encountered the same situation, how did you solve it？
So far, I haven't. Everything I've seen online through searching says I need to change some setting in the unet, but there is no unet for this model because it's a transformer based model, and I have't figured out what the equivalent is in the transformer.