Hi Ray,
That's a new one. I've never seen that before! A quick look at the code
tells me that the only place where askMemory is actually used is in the
functions for reading/writing images from/to disk. Did the mouse ate the
cheese in the expectation or in the maximisation step when it dies? If
it's expectation: I suspect somehow one of your last particles gets read
incorrectly. Are you using a parallel file system? If it's maximisation:
then something is probably wrong in writing the reconstructed reference
images. Do you see something strange in the model.star file in that case?
E.g. zero image size, or so?
HTH,
Sjors
> Dear all,
>
> I've encountered a strange error when using Relion for 3D classification.
> The job ran for days with many iterations, and died (stalled not crashed)
> with an error message:
>
> Error in askMemory: Memory allocation size requested is zero!
> File: src/memory.cpp line: 27
>
> In thread 6
>
> I've tried to resubmit the job, but encountered the same error - just
> died/stalled at different cycle of iterations. However, one thing is
> consistent among the failing runs - they all died at when they were ending
> the iteration while writing files (the Relion mouse almost ate the whole
> cheese!!).
>
> My data set has ~100,000 particles with a image size of 64x64 and a pixel
> size of 4.88. This is probably not an out-of-memory issue - the memory
> usage is supposed to be not that intensive as the stdout at the 33rd
> iteration reported (before my job died):
>
> Estimated memory for expectation step > 1.52443 Gb, available memory =
> 24
> Gb.
> Estimated memory for maximization step > 0.0414566 Gb, available memory =
> 24 Gb.
> Expectation iteration 33 of 50
>
>
> Here is my Relion command line:
> mpirun -map-by node -x OMP_NUM_THREADS -x PATH -x LD_LIBRARY_PATH -n
> $relion_mpi_tasks `which relion_refine_mpi` --o Class3D/run1 --i
> 2D_class_filtered_160420.star --particle_diameter 200 --angpix 4.88 --ref
> CryoComplex_Hsp90_Dmut.mrc --firstiter_cc --ini_high 40 --ctf --iter 50
> --tau2_fudge 4 --K 3 --flatten_solvent --zero_mask --oversampling 1
> --healpix_order 2 --offset_range 8 --offset_step 2 --sym C1 --norm --scale
> --j 8 --memory_per_thread 3
>
>
> Has anyone encountered this? Are there obvious mistakes I've made in the
> command line? Also - are there ways to trigger "--verbose" mode or compile
> a debug mode of RELION?
>
>
> Thanks for the help in advance!
>
> Cheers,
> Ray
>
--
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
|