All model parameters divided into cosmological and HOD parameters, with the training range used by the aemulus emulator and the prior range used in the MCMC fit. Prior ranges were chosen to be slightly larger than the original training ranges, except where excluded by the physical meaning of the parameter, in order to be able to identify if the fit converges outside of the training range. The purpose of this extended range is only to more easily identify a prior dominated fit, since the emulator is not expected to produce accurate clustering outside of the training range. Instead, it would regress to the mean prediction. The exception is log M_cut, where the prior excludes the lower part of training range since log M_cut ceases to have any impact on the halo occupation if it is below log M_min. This is the case for the eBOSS LRG sample, so log M_cut is poorly constrained. However, we found the chains tended to pile up at the lower end of the training range, which gave the misleading impression that the data strongly preferred the lowest possible value, although it had no effect on the cosmological constraints. For that reason, we set a more reasonable lower limit on log M_sat for our sample.