timbrooks

/

instruct-pix2pix

like 838

I want to reproduce the training with the resolution of 512 and based on the pre-trained model sd-v1.5. We train the model 10 epochs. However, our model may generate many poor results.

What could be the cause? Resolution？ The number of training epochs?
And what training settings did you use? (resolution, based model, and training epochs)?
Thx!