This changes the tests to use the new things like t_nounits and D_nounits instead of declaring a parameter t and Differential for every model. This makes it more idiomatic.
Also changed the datafit tests to use the normal Lorenz system instead of the modified one. I think some issues with the CI were being caused because the bayesian_datafits were testing a lot of parameters that made the system difficult to solve, which caused the CI to time out. This also meant I had to give the optimizers a few more timesteps to be in tolerance, but the tests run much faster now.
This also updates to Turing 0.33. I could try to seperate that out, but for some reason when I try to downgrade to 0.30 to test I get precompilation errors that I haven't been able to work out.
Checklist
Additional context
This changes the tests to use the new things like
t_nounits
andD_nounits
instead of declaring a parametert
andDifferential
for every model. This makes it more idiomatic.Also changed the
datafit
tests to use the normal Lorenz system instead of the modified one. I think some issues with the CI were being caused because thebayesian_datafit
s were testing a lot of parameters that made the system difficult to solve, which caused the CI to time out. This also meant I had to give the optimizers a few more timesteps to be in tolerance, but the tests run much faster now.This also updates to Turing 0.33. I could try to seperate that out, but for some reason when I try to downgrade to 0.30 to test I get precompilation errors that I haven't been able to work out.