search

3dem / relion

Image-processing software for cryo-electron microscopy
https://relion.readthedocs.io/en/latest/
GNU General Public License v2.0
450 stars 201 forks source link

Error in 2D classification using GPU not CPU #19

Closed bforsbe closed 7 years ago

bforsbe commented 8 years ago

Originally reported by: Marcus Fislage (Bitbucket: mfislage, GitHub: mfislage)

I was able to successfully import a micrograph STAR file with CTF information, and autopick star and box files. I then was able to extract both unbinned and binned particles. During 2D classification with either set, however, when GPU is enabled I receive the following error after "Estimating initial noise spectra" is completed: "Projector::get2DSlice%%ERROR: Dimension of the data array should be 2 or 3 File: Relion2/install/directory/src/projectorh.h line:224" With CPU only, the job runs fine. Graphics card is GTX970, Cuda is 7.5, and nvidia drivers are 352.39

bforsbe commented 8 years ago

Original comment by Marcus Fislage (Bitbucket: mfislage, GitHub: mfislage):

We downloaded the new git clone, installed local FFTW and FLTK, and recompiled. Now it is working with GPU on a small test set. For a full data set iteration 1 worked and everything seems to be fine.

bforsbe commented 8 years ago

Original comment by Bjoern Forsberg (Bitbucket: bforsbe, GitHub: bforsbe):

I can't seem to get it to fail. Regardless of using --gpu or not, running relion_refine_mpi or just relion_refine, all runs finish without throwing the error you see. I tried both through command-line and GUI. Can you just confirm that running either

#!bash

mpirun -n 3 `which relion_refine_mpi` --o Class2D/job012/run --i Extract/job003/particles.star --dont_combine_weights_via_disc --no_parallel_disc_io --pool 3 --ctf  --iter 25 --tau2_fudge 2 --particle_diameter 200 --K 1 --flatten_solvent  --zero_mask  --strict_highres_exp 15 --oversampling 1 --psi_step 10 --offset_range 5 --offset_step 2 --norm --scale  --j 1 --gpu

or 

#!bash

`which relion_refine` --o Class2D/job012/run --i Extract/job003/particles.star --dont_combine_weights_via_disc --no_parallel_disc_io --pool 3 --ctf  --iter 25 --tau2_fudge 2 --particle_diameter 200 --K 1 --flatten_solvent  --zero_mask  --strict_highres_exp 15 --oversampling 1 --psi_step 10 --offset_range 5 --offset_step 2 --norm --scale  --j 1 --gpu

in the top directory ( 051116 ) does throw the error on your system?

bforsbe commented 8 years ago

Original comment by Marcus Fislage (Bitbucket: mfislage, GitHub: mfislage):

Attached you can find hopefully all necessary files. Just the micrograph is missing. In case you need it as well please tell me.

bforsbe commented 8 years ago

Original comment by Bjoern Forsberg (Bitbucket: bforsbe, GitHub: bforsbe):

I have tried to see where this could be happening, but I think I will need some more information. If you have a minimum example of input files that shows this error that would be ideal. Typically you only need a .star and .mrc(s) file with one or a couple of particles to show these errors, since they occur right away. Would it be possible for you to provide this? That and the command used would allow me to fix it fairly quickly.