Muhammad Rizqi Nur
Muhammad Rizqi Nur
### Is there an existing issue for this? - [X] I have searched the existing issues and checked the recent builds/commits ### What happened? 1. Logging is off by 1...
Providing gradient clipping to user. It's used to handle exploding gradient. I thought it might help for too-large learning rate in hypernetwork training, but it doesn't seem so. Using 0.0001...
1. Added input validations before loading dataset for training. May save time because dataset loading may take a while 2. Added missing info on model log, which are the hash...
Also provide clearer error message for invalid learning rate
Related feature requests: - #3655 - #3415 - #3866 Dropdown to select VAE in Settings tab. You can also use "sd_vae" quicksettings to have it on top like the checkpoint...
~~Simple~~ Fix for #4035. ~~If you'd rather have it to really only keep previous models in the cache dict let me know.~~ Model is now cached when it's about to...
### Is there an existing issue for this? - [X] I have searched the existing issues and checked the recent builds/commits ### What happened? "1" checkpoint caching is useless. When...
There was still some issue or room for improvements in #3986, and some of them are just discovered. 1. The currently loaded VAE from auto isn't reflected in the settings...
Embeddings selected in the embedding blacklist will not be loaded (the embedding dict will be reloaded upon saving settings). Image embeddings are loaded [using the same name as they were...
### Is there an existing issue for this? - [X] I have searched the existing issues and checked the recent builds/commits ### What happened? Webui raises errors when loading extension...