Non-uniform refinement fails

Hi, 3D refinements failed (both, new, and non-uniform) on a dataset with error saying that “all particles should have same contrast”. A screen snapshot is attached. Movies (cryo-sample) were imported in cryosparc and everything was processed in v. 2.15 without errors. 2D classification was run several times to rid of junk. I was able to overcome the error in ‘new’ uniform refinement playing with mini-batch parameters, but non-uniform one refuses to run regardless. Tried on a machine with 64Gb of memory, and on larger machine with 384Gb of core memory. 64Gb machine has a single M6000 GPU with 24Gb memory, the larger one has 4 GPUs with 12 Gb/GPU. How could I locate particles with inverted contrast in a large dataset?

Hi @mbs,

My naive idea is that this error was not caused by your computer hardwares. In my experience, this happens when the particles came from the movies that were processed in separate jobs. Did you process all the movies together, or did you merge several sets of particles from different jobs?


Hi @Feng10, I understand what you’re saying, but that does not explain why different particles would have different contrast. I Even if the jobs were run separately, particles should have same contrast at the end. Another question is, why only 3D refinement complained and none of earlier programs run before that. And the crucial point is how to find offending particles in the set. Actually, I went manually through the images from particles were extracted, and did not see a single one with inverted contrast.
All the best,