You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Brief description: When saving an image in SD.Next, the dynamic thresholding extension is changing the name of the latent sampler in the image's recorded metadata by appending "_dynthresXXX" characters to the end. This is transposed back to latent sampler name field when feeding the image metadata to the Text2Image tab causing it to name a sampler that doesn't exist.
In looking at the code, it appears that when issue #58 was implemented, somehow the thresholding name for the latent sampler is being restored as the latent sampler's original name.
Linking to the bug explanation as originally posted on the SD.Next issues page:
I don't have a fix for this, but I ran a few tests on the code, specifically in dynamic_thresholding.py, lines 163 and 180.
Lines 179-181 correctly restore the names of each sampler, however by that point it appears that the metadata has already been recorded (or copied for later recording) with the altered latent sampler name? It's not clear to me why this same bug doesn't also affect the primary sampler name.
Brief description: When saving an image in SD.Next, the dynamic thresholding extension is changing the name of the latent sampler in the image's recorded metadata by appending "_dynthresXXX" characters to the end. This is transposed back to latent sampler name field when feeding the image metadata to the Text2Image tab causing it to name a sampler that doesn't exist.
In looking at the code, it appears that when issue #58 was implemented, somehow the thresholding name for the latent sampler is being restored as the latent sampler's original name.
Linking to the bug explanation as originally posted on the SD.Next issues page:
vladmandic/automatic#2124
The text was updated successfully, but these errors were encountered: