I would like to use GpABC for my project, however runtimes of my simulator are long enough that I would need to run simulations in parallel on an HPC cluster. If I understand the code and the documentation correctly, at the moment simulations are run strictly sequentially, is that correct?
I would be happy to try and give an implementation a go myself. If I understand the code correctly, the easiest way to do it would probably be to allow for simulator functions that take batches of parameter combinations (and let the simulator take care of parallelisation). For that it seems it would be sufficient to modify simulate_distance and everything upwards from that, or is there anything else to change?
tanhevg
commented
2 years ago
I would like to use GpABC for my project, however runtimes of my simulator are long enough that I would need to run simulations in parallel on an HPC cluster. If I understand the code and the documentation correctly, at the moment simulations are run strictly sequentially, is that correct? I would be happy to try and give an implementation a go myself. If I understand the code correctly, the easiest way to do it would probably be to allow for simulator functions that take batches of parameter combinations (and let the simulator take care of parallelisation). For that it seems it would be sufficient to modify simulate_distance and everything upwards from that, or is there anything else to change?