Closed Rombur closed 2 years ago
Retest this please
The reason the failed is due to the fact that we were unlucky in the choice of the random generator and that we converge to the correct solution only for a large sampling size. We use random number in ensemble_management
and DataAssimilator
. In DataAssimilator
the seed is always the same but in ensemble_management
the seed was different for each run that means that it is impossible to reproduce the result between runs. I have changed that so that the seed is always the same.
Ok -- I'm a little leery of always using the same seed (in both cases really, if I was thinking about it I would have used a random seed in DataAssimilator too) for productions runs, but it should be ok for now.
Yes, the numbers will always be the same between run but they will still be random. It's quite common to fix the seed that way if you change a parameter you know that the difference in the output is due to that parameter.
I was trying to add a new test case but I hit the multithreading bug https://github.com/adamantine-sim/adamantine/issues/130 This PR updates the Docker image and the README