Right now the docs are generated by doing a full fit to the light curve with SMC and NUTS, which is expensive. We should leave in a single example that does this, but move all of the other examples to load a best-fit result that is pre-generated, or use a non-linear optimizer. Maybe we can prepare converged chains, and make a minimal model.pkl, trace.pkl and summary.pkl file out of them?
Right now the docs are generated by doing a full fit to the light curve with SMC and NUTS, which is expensive. We should leave in a single example that does this, but move all of the other examples to load a best-fit result that is pre-generated, or use a non-linear optimizer. Maybe we can prepare converged chains, and make a minimal
model.pkl,trace.pklandsummary.pklfile out of them?