Hi,
> I am unsure of how to approach this because we have run polishing successfully on our cluster with a very similar dataset (134,000 particles)
> containing the same imaging parameters,
What matters is not the total number of particles but the number of particles per movie.
Memory consumption also depends on the number of frames per movie, the box size (before down-sampling).
You can try:
- reduce the number of MPI process per node
- use the --sbs option (but slower)
- rebuild with CPU single precision (cmake -DDoublePrec_CPU=OFF)
- use the --pad 1 option
Best regards,
Takanori Nakane
On 2021/02/11 19:28, Matthew Doran wrote:
> Hello,
>
> In relion/3.1.1 I have trained a dataset for polishing and am attempting to run the polishing step with optimized parameters. I am running this on a
> data set of filaments with 144,000 particles at a pixel size of 1.058A (binned 2x during motioncor from .529A).
>
> During the + Performing loop over micrographs ... portion of the output I get this error:
>
> 0.20/3.96 hrs ...~~(,_,">in: /scratch/522357.1.budge/relion/src/multidim_array.h, line 743
> ERROR:
> Allocate: No space left
> - Warning: unable to load raw movie frames for MotionCorr/job001/movies/Actin_331_00375_Aug27_04_13_00.mrc. Possible
> reasons include lack of the metadata STAR file, the gain reference and/or the movie.
> in: /scratch/522357.1.budge/relion/src/multidim_array.h, line 743
> ERROR:
> Allocate: No space left
> - Warning: unable to load raw movie frames for MotionCorr/job001/movies/Actin_331_00196_Aug27_02_05_58.mrc. Possible
> reasons include lack of the metadata STAR file, the gain reference and/or the movie.
> in: /scratch/522357.1.budge/relion/src/multidim_array.h, line 743
>
> This error continues listing more movie files until our cluster auto-kills the job. My initial command is:
>
> mpirun -np 7 --map-by socket --bind-to socket `which relion_motion_refine_mpi` --i CtfRefine/job002/particles_ctf_refine.star --f
> PostProcess/job022/postprocess.star --corr_mic MotionCorr/job001/corrected_micrographs.star --first_frame 1 --last_frame -1 --o Polish/job003/
> --params_file Polish/job001/opt_params_all_groups.txt --combine_frames --bfac_minfreq 20 --bfac_maxfreq -1 --j 14 --pipeline_control Polish/job003/
>
> I am unsure of how to approach this because we have run polishing successfully on our cluster with a very similar dataset (134,000 particles)
> containing the same imaging parameters, so the memory requirements shouldn't have changed dramatically.
>
> Any help would be greatly appreciated!
>
> --
> Matt Doran
> PhD Candidate
> Department of Biophysics and Physiology
>
>
> ------------------------------------------------------------------------------------------------------------------------------------------------------
>
> To unsubscribe from the CCPEM list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1 <https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1>
>
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1
This message was issued to members of www.jiscmail.ac.uk/CCPEM, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|