by Martin KavalarOct 29 2018
Making science reproducible @nextjournal

Relion Benchmark

This article reproduces the relion benchmarks. They should be compared to this reference.

We start by downloading the benchmark data. This is a 50GB file. Once successfully downloaded, we've locked the cell so it doesn't run again.

wget --progress=bar:force -P /results

Let's check we have all the data:

Let's see what kind of GPU's we're using:


And memory:

awk '/MemFree/ { printf "%.3f \n", $2/1024/1024 }' /proc/meminfo

Next, we unpack the data.

Then we start a benchmark run:

cd /relion_benchmark/ && mpirun --allow-run-as-root -n 2 `which relion_refine_mpi` --i Particles/ --ref --firstiter_cc --ini_high 60 --ctf --ctf_corrected_ref --iter 25 --tau2_fudge 4 --particle_diameter 360 --K 6 --flatten_solvent --zero_mask --oversampling 1 --healpix_order 2 --offset_range 5 --offset_step 2 --sym C1 --norm --scale --random_seed 0 --o ../ --gpu 0 --pool 100 --dont_combine_weights_via_disc --j 5