Model move to device - gpu

distributed
1

In the documentation,

It is mentioned that the tensors must be assigned to a new variable when using to device.
mytensor = my_tensor.to(device)

Should we not do the same for the model?

Like,
model = model.to(device)

@sungkim11 @fritzo @neerajprad @alicanb @vishwakftw

2

.to() modifies the model in-place.