Huw Jenkins

OK I can reproduce with CUDA-11.3.1 on a GPU node on our cluster (V100 cards) so can more easily roll back CUDA versions now. I just have to wait in...

OS: CentOS Linux release 7.9.2009 GCC: 8.3.0 CUDA: 10.1 RELION: 4.0-beta-2-commit-9b23e5 ``` loop_ _rlnCoordinateX #1 _rlnCoordinateY #2 _rlnClassNumber #3 _rlnAutopickFigureOfMerit #4 _rlnAnglePsi #5 846.617652 180.611766 4 2.630840 40.000000 304.782355 440.241179...

OS: CentOS Linux release 7.9.2009 GCC: 8.3.0 CUDA: 11.7.0 RELION: 4.0-beta-2-commit-9b23e5 ``` loop_ _rlnCoordinateX #1 _rlnCoordinateY #2 _rlnClassNumber #3 _rlnAutopickFigureOfMerit #4 _rlnAnglePsi #5 2585.005898 1839.982364 3 2.669658e+06 135.000000 1794.829422 3510.641197...

Thanks for confirming it doesn't cause any problems! Fortunately the warning is only printed the first time it happens.

Further fun(?). Using `combine.Imid=0`: ``` Performing multi-dataset profile/summation intensity optimisation. +---------------+---------+---------+ | Combination | CC1/2 | Rmeas | |---------------+---------+---------| | prf only | 0.99937 | 0.03105 | +---------------+---------+---------+ Profile intensities...

Thanks for looking into this! > in the default mode, dials.scale will use profile estimates up until the point of intensity combination, and then the best estimate determined by the...

I have been investigating this further. I followed the steps in the small molecule tutorial data (choosing to integrate in P1) and then ran: ``` dials.symmetry integrated.refl integrated.expt ``` to...

If I repeat this with `minimisation=regression` then I get much more similar results for each of the 20 sets of reflections: ``` remove_0 a = 1.21277, b = 0.04041 remove_1...

That seems like an incredible number of strong pixels? Are the spot finding parameters appropriate for these images?

ok fair enough! I couldn't replicate as DIALS imported those images as 18 sweeps and extracted the spots per sweep: ``` Finding spots in image 100 to 100... Setting nproc=10...