Especially but not only for training, it would be very useful to have a scheduler that undertakes evaluations (be it EFV or even basis etc) for multiple systems at once. E.g. 100 small systems can be converted into a disconnected graph and all evaluated in parallel on GPU, or even on CPU this can probably leverage SIMD better.
Especially but not only for training, it would be very useful to have a scheduler that undertakes evaluations (be it EFV or even basis etc) for multiple systems at once. E.g. 100 small systems can be converted into a disconnected graph and all evaluated in parallel on GPU, or even on CPU this can probably leverage SIMD better.