stable-diffusion how to train with fp16? when i train with fp16,got an error: expected scalar type Half but found Float

how to train with fp16? when i train with fp16,got an error: expected scalar type Half but found Float

Open AIXiaoBaiDemon opened this issue 2 years ago • 3 comments

Nov 08 '22 15:11 AIXiaoBaiDemon

What model are you trying to train? Is this a variations model or normal stable diffusion?

Dec 09 '22 14:12 justinpinkney

I want to train with normal stable diffusion.

Dec 12 '22 01:12 AIXiaoBaiDemon

Is the problem solved? I also encountered the same problem.

May 12 '24 13:05 yisunhan