The GPUlayer Trainer runs the same PyTorch script in your process, on every GPU of a machine, or across many machines: laptops, workstations, servers and cloud VMs. It picks machines and GPUs that can train the model, starts every process, and brings back logs, checkpoints and metrics.