Multiple GPU about shap-e

[notpatrickadams]

According to this StackOverflow answer, it looks like using cuda uses all available GPUs

[notpatrickadams]

There are some examples provided in the repository

[xaranex1]

How can i used it in code? Can you write some example for jupyter notebook?

[xaranex1]

You think nn in models? So if i put nn instead of nerf it will be work?

[xaranex1]

I dont know how can i do it, jupyter notebooks examples are without torch.nn

[xaranex1]

I tried to modified code like this:

xm = torch.nn.DataParallel(load_model('transmitter', device=device))
model = torch.nn.DataParallel(load_model('text300M', device=device))

but i have error: AttributeError: 'DataParallel' object has no attribute 'd_latent'

[xaranex1]

Is there some solution?

[xaranex1]

How can i use multiple GPUs ?