Description of Supplementary Data This record contains the samples used to create the figures (excluding validation and prior dependence plots) and to derive most of the results in Hoof et al., “Axion global fits with Peccei-Quinn symmetry breaking before inflation using GAMBIT” (available on the arXiv). Please contact the authors if you are interested in other samples, YAML files or plotting scripts. This record consists of 21 YAML files (6 for T-Walk, 15 for Diver). Running ./gambit -f path/to/YAML/file.yaml in the GAMBIT directory will start the scan. However, most users might want to adjust the output file name and directory as well as the settings for the samplers to their systems. 21 hdf5 files (6 for T-Walk, 15 for Diver). These files contain the actual samples and were compressed using the tar format. Two example pip files (2_QCDAxion_10M1.pip for Diver samples, 2_QCDAxion_3041.pip for T-Walk samples) for producing plots from the corresponding hdf5 files, using pippi and functions.py. The files follow the naming scheme V_ModelName_[S][C][I][R][E] plus one of the extensions .yaml, .hdf5.tar.gz, or .pip. V: This internal version number can be ignored, but should be quoted when asking for help with the plotting scripts ModelName: Corresponds to the axion models in the paper (GeneralALP, QCDAxion, DFSZAxion_I, DFSZAxion_II, KSVZAxion) S: Scanner (S=1: Diver, S=3: T-Walk) C: Switch to include (C=1) or exclude (C=0) the White Dwarf cooling hints I: Setting for the initial misalignment angle θi (I=4: flat prior on θi with values in [-3.1415, 3.1415]). I=M is used to indicate that the file includes merged samples from other scans in addition to the corresponding I=4 scan. R: Setting for the DM relic density likelihood (R=1: upper limit, R=2: matching the DM density) E: Extra digit for the anomaly ratio E/N; only for KSVZAxion models (E=1: 0, E=2, 2/3, E=3: 5/3, E=4: 8/3), DFSZAxion-I models (E=1: 8/3), DFSZAxion-II models (E=2: 2/3), or some GeneralALP files (E=a: “QCD-like setting” with β = 7.94, Tcrit = 147 MeV; E=b: “Simple ALP-like setting” with β = 0, Tcrit irrelevant) For convenience, we provide a mapping between the figures in the paper and the hdf5 files: Fig. 1: none Figs 2 - 11: Validation plots Figs 12 + 13: 2_GeneralALP_10M2 Fig. 14: 2_GeneralALP_10M2a, 2_GeneralALP_10M2b Fig. 15: 2_QCDAxion_10M1, 2_QCDAxion_10M2 Fig. 16: 2_QCDAxion_3041, 2_QCDAxion_3042 Figs 17 + 18: 2_QCDAxion_10M1, 2_QCDAxion_10M2, 2_QCDAxion_30M1, 2_QCDAxion_30M2 Fig. 19: 3_KSVZAxion_10M11, 3_KSVZAxion_10M12, 3_KSVZAxion_10M13, 3_KSVZAxion_10M14, 3_DFSZAxion_I_10M11, 3_DFSZAxion_II_10M12 Fig. 20: 2_QCDAxion_10M1, 3_KSVZAxion_10M11, 3_KSVZAxion_10M12, 3_KSVZAxion_10M13, 3_KSVZAxion_10M14, 3_DFSZAxion_I_10M11, 3_DFSZAxion_II_10M12 Fig. 21: 2_QCDAxion_11M1, 2_QCDAxion_11M2 Fig. 22: 2_QCDAxion_3141, 2_QCDAxion_3142 Figs 23 + 24: 2_QCDAxion_3041, 2_QCDAxion_3042, 2_QCDAxion_3141, 2_QCDAxion_3142 Fig. 25: 2_QCDAxion_3141, 2_QCDAxion_3142 Fig. 26: 2_QCDAxion_11M1, 3_DFSZAxion_I_11M11, 3_DFSZAxion_II_11M12 Fig. 27: 2_QCDAxion_3141, 3_DFSZAxion_I_31411, 3_DFSZAxion_II_31412 Fig. 28: Validation plot Fig. 29: Prior dependence plot A few caveats to keep in mind: The YAML files are designed to work with GAMBIT 1.3.0, and the pip files are tested with pippi 2.1, commit 1a08644. They may or may not work with later versions of either software (these working versions/commits can always be obtained via the git history). The pip files will produce an approximately complete, but very basic version of plots in the paper. Re-creating all the plots in the paper requires various manual, undocumented interventions such as additions, deletions and combination of the plotting scripts created by pippi. Users wishing to reproduce the more advanced plots in the paper should contact the authors for tips, scripts, or experiment for themselves.