WeatherDiffusion
WeatherDiffusion copied to clipboard
IGITUGraz
About training
Dear author I'd like to ask you three questions
Most importantly, the training code seems to be incomplete (for example, the training process does not match the description in the paper, especially regarding patch). Is there a more sound training code? Thank you very much.
Second, are data_transform and inverse_data_transform necessary?
Finally, when I was training 128*128 pictures on Nvidia 3090, a single card could actually run the batch of 24 data. Is this normal?
Most importantly, the training code seems to be incomplete (for example, the training process does not match the description in the paper, especially regarding patch).
The code is complete and works fine. This repository accurately represents the method described in the paper. I do not understand what is the problem exactly, maybe you should specify which part from the paper seems to not match?
Second, are data_transform and inverse_data_transform necessary?
Yes, for the current implementation version they are necessary.
Finally, when I was training 128*128 pictures on Nvidia 3090, a single card could actually run the batch of 24 data. Is this normal?
I am not sure regarding the specifics of that particular GPU, so this is a matter of trying out different batch sizes and seeing which works suitably on your machine.
