ZT comments

Results 18 comments of

ZT

[Reproduction issue] Semantic image synthesis and layout-to-image cannot be reproduced

Also waiting for the semantic synthesis training pipeline

Semantic Image Synthesis

Hi, does anyone know how to train this semantic synthesis model?

Hyper parameter intuition

check the original paper from NVIDIA on elucidated imagen:https://arxiv.org/pdf/2206.00364.pdf section E2

video synthesis using elucidatedImagen

> From what it looks like, you're not training the second Unet. You have to make separate calls to `trainer.train_step(unet_number = 2 max_batch_size = 4)` to train the up-scaling unet...

Loss oscillation

I got a similar loss curve. Did you observe any visual improvement in your synthesized image?

I trained my model to synthesize gray scale images (channels = 1). This is what it looks like after 5000 steps: ![Screen Shot 2022-10-20 at 10 21 58](https://user-images.githubusercontent.com/25071118/196909884-ceacf439-92ed-4194-93a9-bddd142a381d.jpg) This is...