stabilityai

/

stable-diffusion-3-medium

like 3.04k

Load the basic workflow and run.
When the K Sampler is running, an error occurs at line 937 （ x = self.x_embedder(x) + self.cropped_pos_embed(hw, device=x.device).to(dtype=x.dtype)
）of mmdit.py as follows:
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!