Reliability of OccamNet Tests

SciML / DataDrivenDiffEq.jl

Data driven modeling and automated discovery of dynamical systems for the SciML Scientific Machine Learning organization

https://docs.sciml.ai/DataDrivenDiffEq/stable/

MIT License

402 stars 56 forks source link

Reliability of OccamNet Tests #328

Open AlCap23 opened 2 years ago

AlCap23 commented 2 years ago

It seems there are some issues with the tests of OccamNet since recently, probably related to infeasible combinations of functions ( e.g. 1 / 0, log(-10.0) or similar calls ) . This should be a reasonable quick fix, but I open this issue as a reminder and because I will probably have time for next week.

AlCap23 commented 2 years ago

Just some tasks related

[ ] Add more tests to src\symbolic_regression/occamnet.jl, see here
[ ] I am running 100 training loops locally right now. There are no errors using v1.7.1 on the current MacOs. Why do the tests fail on the CI