train model for different $w$

[betterze]

Thank you for your great work. I have a simple question.

In the training section, you mention that 'change repleated $w$ for all layers to different $w$s'. There are a few pretrained models in stylegan which use repleated $w$, could I just adapt the weights from them or I need to retrain the model?

Thank you for your help.

[betterze]

And would you mind uploading the manipulation boundary for face? Thank you again for your help.

[betterze]

any response?

[ShenYujun]

And would you mind uploading the manipulation boundary for face? Thank you again for your help.

Sorry, but the npy boundaries are ignored by git before. Now, all boundaries are uploaded.

[ShenYujun]

Thank you for your great work. I have a simple question.

In the training section, you mention that 'change repleated $w$ for all layers to different $w$s'. There are a few pretrained models in stylegan which use repleated $w$, could I just adapt the weights from them or I need to retrain the model?

Thank you for your help.

Although we recommend using different ws for different layers, using same w (i.e., the official StyleGAN) also works. You can use the official StyleGAN model and then use IDInvert to train your own encoder and use InterFaceGAN to find your own boundaries.

[betterze]

And would you mind uploading the manipulation boundary for face? Thank you again for your help.

Sorry, but the npy boundaries are ignored by git before. Now, all boundaries are uploaded.

Thanks a lot, I will try it.