Looks like that works! Any additional input for how to make a dictionary work in conjunction with dataParallel to use multiple GPUs? I’m now getting “RuntimeError: The size of tensor a (8) must match the size of tensor b (16) at non-singleton dimension 0” Which I’m guessing is because some values are being split across GPUs and others are not.
asked in a new thread here: Resetting Dataparallel after initialization, and using it in a dictionary