I have now successfully run the new tract mode FGCM from DM-16704 on the 3 tracts from RC2 (9615, 9697, 9813). The fgcm_photoCalib files are on lsst-dev01 in a rerun at /datasets/hsc/repo/rerun/private/erykoff/fgcm_tract_w_2019_30/fgcmtract-run4.

The runtime was approximately 30 minutes for each of the tracts (calibrating all of grizy simultaneously), and approximately 70-75% of that time was spent reading in the calexp sources.

I have attached some plots here demonstrating the internal metrics. I think I need to run Lauren MacArthur's visit check code as independent validation, and I'd also love to do coadds and have that validated as well.

The following table summarizes the important numbers. "int" refers to internal star repeatability in mmag. These are for bright (s/n > 100) reserved stars (not used in the fit), and are after subtracting out the typical photometric error. "ref" refers to reference star consistency in mmag. These are for bright-ish (s/n > 50) PS1 reference stars, and are the total value (e.g., does not subtract off photometric errors).

The plots also show these broken down by the 25%/50%/25% bluest, middle, and reddest stars.

Do let me know if you would like me to run the visitAnalysis scripts on your outputs (assuming they are on lsst-dev, ah, yes, they are!).  I offer in part as I would be using a not-fully-committed branch for DM-19189 that implements the S/N-based selection for computing metrics that I'm sure you would prefer (to the current mag-based selection on master).  The mutual benefit might work as I do plan to ask you to review that ticket, and any outputs I produce for you here could be the example output for you to view as reviewer 

Lauren MacArthur Yes they are! In order to read them, you need the DM-16704 branches of obs_base and obs_subaru which define the fgcm_tract_photoCalib dataset. But this should be fully compatible with jointcal_photoCalib, with the same dataId requirements.
I definitely think we want to do this with the S/N based selection, and I'm happy to review that ticket when it's ready. Thanks!

Ah, right, I'll need to update the scripts in pipe_analysis to have the option of applying your fgcm photometric calibrations in lieu of the jointcal ones. What should be done about astrometric calibration in this case (can the jointcal ones be used)? I'll go ahead and make a ticket for that.

I think you should be able to use the jointcal astrometric calibration, or even the original from the calexps. And I see that the ticket is DM-21052, thanks!

The fgcmcal code has successfully run, so I'm closing this ticket. If there are downstream issues we can file more tickets for improvements.