| Age | Commit message (Collapse) | Author |
|---|
|
Allow trailing comma in learning rate
|
|
|
|
prototype progress api
|
|
value.
|
|
Save Hypernetwork strength to infotext.
|
|
|
|
|
|
|
|
3825-save-hypernet-strength-to-info
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[API][Feature] Add png info endpoint
|
|
|
|
|
|
|
|
|
|
|
|
for #3768
|
|
|
|
|
|
|
|
Reduce peak memory usage when changing models
|
|
|
|
|
|
Revamped Arabic localization
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Natural sorting for dropdown checkpoint list
|
|
Fix log off by 1 #3847
|
|
Fixed proper dataset shuffling
|
|
Added option to use unmasked conditioning image for inpainting model.
|
|
Extras Tab - Option to upscale before face fix, caching improvements
|
|
|
|
Filename tags are wrongly referencing to process size instead of image size
|
|
Add missing support for linear activation in hypernetwork
|
|
Disable unavailable or duplicate options for Activation functions
|
|
|
|
|
|
Filename tags [height] and [width] are wrongly referencing to process size instead of resulting image size. Making all upscale files named wrongly.
|
|
|
|
|
|
This changes the extras image cache into a Least-Recently-Used
cache. This allows more experimentation with different upscalers
without missing the cache.
Max cache size is increased to 5 and is cleared on source image
update.
|
