note you should also use trace.to_netcdf() not trace.posterior.to_netcdf (with this you’ll only save the posterior samples, and ArviZ expects data to be in groups: posterior for posterior samples, sample_stats for sampling info and diagnostics like divergences…
2 Likes