DeepH-pack icon indicating copy to clipboard operation
DeepH-pack copied to clipboard
verified publisher icon mzjb

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

Open Lonya0 opened this issue 1 year ago • 1 comments

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? 1725765332852

Lonya0 avatar Sep 08 '24 03:09 Lonya0

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

mzjb avatar Nov 04 '24 07:11 mzjb