Comment on Stable Diffusion forming images from text: image snapshots at each stepparentComments−beecafe4yYou don't even need to do a diff, the model itself actually predicts the diff (which is how it samples the image) so you could just stop once the model is predicting close to 0−stavros4yIs there a parameter for this, to stop when subsequent iterations do nothing?
Comments
You don't even need to do a diff, the model itself actually predicts the diff (which is how it samples the image) so you could just stop once the model is predicting close to 0
Is there a parameter for this, to stop when subsequent iterations do nothing?