train 13000 step, texture is much different from reference image

[VisionU]

In first stage, In oder to train model easy, we remove posenet, use VAE encoder instead（cat with noise like in-painting SD). Now it train 13000 step， texture is much different from reference image, is this correct? or we just need to waiting for more steps?

[guoqincode]

I don't think it's normal... It looks scary.

[guoqincode]

I will release the pre-trained model as soon as possible, please be patient, it may not take a few days.

[VisionU]

I will release the pre-trained model as soon as possible, please be patient, it may not take a few days. @guoqincode OK, I train it use 512x768, use r.clear(), w.clear(), can you give some advise for this bad result? data, or code?

[guoqincode]

If it's a texture issue, I have two suggestions.

1. Train more steps
2. Check if unet is updated

[VisionU]

If it's a texture issue, I have two suggestions.

1. Train more steps
2. Check if unet is updated

Thank you ~

[sizhky]

Check if unet is updated Can you elaborate @guoqincode

[renrenzsbbb]

If it's a texture issue, I have two suggestions.

1. Train more steps
2. Check if unet is updated

Thank you ~

Can you fix it? If can, can you give some advices?