At the moment when running replicas in parallel, the initial state of the network depends on the number of replicas (i.e., it is not the same to run 1 to 20 than doing two runs of 1 to 10, 10 to 20).
While GPU reproducibility is difficult to achieve, the initial state should at least (even if it needs to be done artificially) forced to be the same.
See #1881 (comment)
(note that all released fits are run sequentially)
At the moment when running replicas in parallel, the initial state of the network depends on the number of replicas (i.e., it is not the same to run 1 to 20 than doing two runs of 1 to 10, 10 to 20).
While GPU reproducibility is difficult to achieve, the initial state should at least (even if it needs to be done artificially) forced to be the same.
See #1881 (comment)
(note that all released fits are run sequentially)