 |
 |
Hi John, I assume you mean release 2.0.x. The first iteration is probably based on getting maximum cross-correlation coefficients rather than integrating over all probabilities, so that's much faster. In the second you then generate seeds for the classification, and the third iteration is the first 'real' one. From there on it should get faster again. If you have enough RAM you might consider running 5 or 7 MPIs each of them with correspondingly fewer threads. That may be a bit more efficient. But, running on only 32 CPUs will always be slow.Can you run across 4 nodes? Or better perhaps, can you run on GPUs instead? HTH, Sjors On 05/22/2017 09:07 PM, John Heumann wrote: > Can anyone give me some suggestions on what might cause the first exhaustive 3D classification to run extremely slowly under Relion 2.2? Specifically, I've got ~50,000 particles (box size 128) from ~280 micrographs, an initiial model low pass filtered to 5 nm, and am trying to classify into 4 classes using 25 iterations, as in the 2.2 tutorial. I'm running on a Linux cluster, with a dedicated single node (32 real cores, 32 hyper-threaded) using 3 MPI process and 20 threads per process. The first iteration only took about 20 minutes. The 2nd took about 90 minutes, and the 3d took 27 hours. At this rate, there's no telling how long all 25 iterations might take. I should add that there's no sign of thrashing or anything abnormal on the node running this job. Top output looks normal, and over the day and half it's been running so far, I've used up 1070 hours of cup time... i.e. about 40X wall time, which is about what I'd expect. > > On a related note, what's the best way to proceed if in the middle of a single particle analysis I decide to remove certain particles or micrographs after picking / extraction / sorting have already been done? Obviously, I can manually edit files, but it seem likely there's a better way through the gui? > > Thanks! > -jh- -- Sjors Scheres MRC Laboratory of Molecular Biology Francis Crick Avenue, Cambridge Biomedical Campus Cambridge CB2 0QH, U.K. tel: +44 (0)1223 267061 http://www2.mrc-lmb.cam.ac.uk/groups/scheres