error: cuda out of memory. when training a YIG model

[Lonya0]

Using nvidia rtx 4090 (24G memory), to train a YIG model, whose hamiltonian.h5 file sizes 230MB, encountered error: cuda out of memory. I have go through CrI3 test and successed before. How can I make it possible to train this model?

[mzjb]

You could try switching to DeepH-E3 for training. It has the same data interface as DeepH but requires less GPU memory.