search

SciML / StochasticDiffEq.jl

Solvers for stochastic differential equations which connect with the scientific machine learning (SciML) ecosystem
Other
248 stars 66 forks source link

CI for LTS #445

Closed ChrisRackauckas closed 2 years ago

ChrisRackauckas commented 2 years ago

@frankschae this is the last one I've been working on.

ChrisRackauckas commented 2 years ago

We might need to move some tests to the buildkite, but let's see if this is ready to merge first 😅

ChrisRackauckas commented 2 years ago

Some of those test failures on v1.7 look way off?

frankschae commented 2 years ago

Yes... but I think the errors have already alternated at least once (i.e. once it failed at 1.6, then at 1.7) when increasing the number of trajectories. The number of trajectories is still on the lower end. Roessler uses 1e8 to 1e9 (https://link.springer.com/article/10.1007/s10543-007-0130-3)

ChrisRackauckas commented 2 years ago

I'd say, swap all of those big weakconvergence tests to buildkite right now. This might take awhile longer than I had originally thought 😅

frankschae commented 2 years ago

The tests, where I've just increased the number of trajectories, were still relatively small up to now (except the 1e7 ... that one is indeed a little odd ..). Basically, we haven't seen those before because weak_convergence2 was failing and the other tests were canceled and didn't run through.

frankschae commented 2 years ago

I'd say, swap all of those big weakconvergence tests to buildkite right now. This might take awhile longer than I had originally thought 😅

yup, good idea. I'll do it.

ChrisRackauckas commented 2 years ago

oops, looks like this is done but a regression on master needs to be fixed. Good time to have the tests working. Thanks @frankschae 😅