Home

Welcome to the Hypernetwork-MonkeyPatch-Extension wiki!

Here, I will introduce how to 'tune' Hypernetworks with extension.

First, in Training tab, you can save template for your hypernetworks.

At default, the .json file will be saved at stable-diffusion-webui\models\hypernetworks.

Second, in Train_gamma tab, you can save file, same way as hypernetwork template.

The 'check' means the values will be saved. 'Cross' means values are not saved.

Train_Tuning tab

Hypernetwork selecting is optional.

Hypernetwork name is required, since hypernetworks will be automatically created from given name.

Newly created hypernetworks will have _i_j suffix.

i is index of current hypernetwork template.

j is index of current training template.

However, you can select hypernetwork and leave it empty. Then it will copy the selected hypernetwork, and train further from there.

Specially, you can specify 'dataset seed'. This fixes Random Number Generator that controls dataset orders.

By this setting, you can compare between settings accurately.

I strongly recommend tensorboard logging feature, which makes you able to compare between multiple settings and results.