I run the code of inference/text_to_image.ipynb, and get speckled results.
text prompt: Epic anime artwork of a wizard atop a mountain at night casting a cosmic spell into the dark sky that says 'Stable Diffusion 3'.
results:
global
local
This appears to be an issue with precision. However, I've adjusted all precisions in the code to torch.float32 and switched to the corresponding model, but the problem persists. What should I do to get normal results? My runtime environment is consistent with the requirements.txt.
I run the code of
inference/text_to_image.ipynb
, and get speckled results.text prompt:
Epic anime artwork of a wizard atop a mountain at night casting a cosmic spell into the dark sky that says 'Stable Diffusion 3'.
results:
This appears to be an issue with precision. However, I've adjusted all precisions in the code to
torch.float32
and switched to the corresponding model, but the problem persists. What should I do to get normal results? My runtime environment is consistent with therequirements.txt
.