I’ll leave this here for anyone who stumbles across the same issue I had:
If using super resolution EER data, that is, EER upsampling 2 during Import in CryoSPARC, when running polishing in RELION after conversion from CryoSPARC, you must add the --eer_upsampling 2 flag in “Additional arguments” for Bayesian polishing.
RELION will complain, but without it you will get a gain reference error crash (see bottom for example).
Further, you will also need to add --eer_grouping XYZ (where XYZ is the closest number of frames to generate the number of frames requested during Import in CryoSPARC) to Additional arguments. If you are using multiple datasets, you must either polish them independently, or work with the lowest common denominator for total frames and live with potential drift in motion grouping on the movies with higher EER frame counts - I’ve no evidence yet whether that is a complete gamebreaker or not, but motion correction is going insanely slow so it will be some time before I know one way or the other.
Again, RELION will complain, but without it, polishing would crash.
If you were crazy enough to use EER upsampling 4 during Import in CryoSPARC, polishing in RELION won’t work at all because RELION no longer supports 16K sampling (and it was buggy when it did).
Gain error:
Size of input gain: X = 4096 Y = 4096 Expected: X = 2048 Y = 2048
in: /home/cryo/build/relion/src/renderEER.h, line 169
ERROR:
Unsupported gain size in EERRenderer::loadEERGain()
- Warning: unable to load raw movie frames for J6/motioncorrected/009205363351292283250_FoilHole_22313357_Data_19218937_19218939_20230722_155434_EER_patch_aligned_doseweighted.mrc. Possible reasons include lack of the metadata STAR file, the gain reference and/or the movie.
Grouping error is too long for reasonable inclusion.