kligvasser / singan Goto Github PK

Hello, I trained a model with your code, and then tested it with your code. I found that the effect in the generated image is very poor. Is there a problem?

is--model-to-load and --amps-to-load when testing image?
where is generative model and dicriminate model located when testing a image?

Hi,
I'm testing your SinGAN implementation with PyTorch 2.01 (using the pytorch/pytorch:latest docker image) and seems working nicely.
The is only a Warning

/opt/conda/lib/python3.10/site-packages/torch/optim/lr_scheduler.py:152: UserWarning: The epoch parameter in `scheduler.step()` was not necessary and is being deprecated where possible. Please use `scheduler.step()` to step the scheduler. During the deprecation, if epoch is different from None, the closed form is used instead of the new chainable form, where available. Please open an issue if you are unable to replicate your use case: https://github.com/pytorch/pytorch/issues/new/choose.
  warnings.warn(EPOCH_DEPRECATION_WARNING, UserWarning)

is not clear to me if it refers to your code or something internal to Torch itself but I think that you would be interested knowing.

Hello! Thank you for putting the work in to:
A) clean up the original SinGAN
B) Keep it up to date with the lastest python and cuda and torch

I'm currently working through the python code as well as the original SinGAN code, and trying to see how to replicate some of the more user-friendly python scripts that allowed us non-AI-researchers use this GAN for paint2Image, Editing, Harmonizing, etc.

Are there any plans to support more detailed scripts like this, and if not, would PR's be accepted If I put the work into replicating their functionality?