How to run Nvidia Modulus v22.09 with FP64

I wish to ask how to run Nvidia Modulus v22.09 with FP64 (double precision).
I did not find the way to set this double precision in official tutorial. I am using Nvidia A100 GPU accelerator.

Hi @Spartacus

We have some defaults defined in the constants file:
https://gitlab.com/nvidia/modulus/modulus/-/blob/release_22.09/modulus/constants.py#L17

This should update Modulus datasets / dataloaders to use FP64. I would also use the PyTorch way to update the default tensor type torch.set_default_tensor_type(t) which should update models to FP64.