Closed brantfaircloth closed 4 years ago
Just to answer the ancillary part of my question about the "missing" files, in case anyone hits this issue and needs a workaround: I just started the job back up from the *.ckp
file and specified 50 bootstrap replicates (see below). This created the desired files.
#!/bin/bash
#PBS -q checkpt
#PBS -A hpc_allbirds03
#PBS -l walltime=72:00:00
#PBS -l nodes=13:ppn=20
#PBS -V
#PBS -N raxmlng-std-mpi
#PBS -o raxmlng-std-mpi.out
#PBS -e raxmlng-std-mpi.err
module load gcc/6.4.0
module load impi/2018.0.128
cd $PBS_O_WORKDIR
SEED=$RANDOM
echo $SEED
mpiexec -np 260 -machinefile $PBS_NODEFILE /project/brant/shared/bin/raxml-ng-mpi \
--msa 75p_alignments.phylip.raxml.rba \
--seed $SEED \
--all \
--bs-trees 50
Thanks for reporting! This should be fixed now with the last commit (https://github.com/amkozlov/raxml-ng/commit/66ad9d22333263f6d1d3643328bd3bed8721dc6a). Could you please confirm?
Closing, please feel free to reopen if the problem persists.
We've observed a problem in 0.9.0 where running autoMRE with MPI seems to hang once 50 bootstrap replicates are generated (and the code would typically evaluate them for convergence and then move on if not converged).
An ancillary issue that arises with this problem is that neither the bootstrap replicates nor the best ML tree files are visible when the run dies. Is there a way to extract these from the
*.ckp
files?We've been calling RAxML pretty simply (so far - e.g. no hybrid mode) within the following:
Thanks very much, b