stabilityai

/

stable-cascade

like 1.2k

How to determine what resolutions a neuron can generate?
For example, I cannot generate images in the format:

• 1200, 632 - 400: pixel_unshuffle expects width to be divisible by downscale_factor, but input.size(-1)=309 is not divisible by 2
• 856, 320 - 400: pixel_unshuffle expects height to be divisible by downscale_factor, but input.size(-2)=85 is not divisible by 2