Multi-machine inference with PyTorch

If you want to use model sharding, this simple example might be useful.
The linked tutorial explains a distributed setup, so let me know, if I misunderstood your use case.