Problem
Optimized reconstructions on the CPU can be slow, early testing showed that we get significantly improved performance by running them on the GPUs
Proposed solution
Thankfully, since we're just using PyTorch, we need to simply adjust some of our to_device calls, and create certain arrays on the GPU. (there's probably more to it than that, and I do expect some edge cases)
Alternatives you have considered, if any
Additional context
Note relevant experimental conditions or datasets
Problem
Optimized reconstructions on the CPU can be slow, early testing showed that we get significantly improved performance by running them on the GPUs
Proposed solution
Thankfully, since we're just using PyTorch, we need to simply adjust some of our to_device calls, and create certain arrays on the GPU. (there's probably more to it than that, and I do expect some edge cases)
Alternatives you have considered, if any
Additional context
Note relevant experimental conditions or datasets