I’m currently debugging an issue in a model I’m working on where some tensors aren’t properly being sent to the GPU. I run the model remotely on a machine with a GPU, but develop and debug the model locally on my laptop without a GPU. I can’t recreate the issue locally because everything it sent to the CPU since no GPU is available.
Is there a way to create an alias of the CPU as a dummy device that I can send my tensors to in order to debug my issue?