[AMBER] Problem Running E-REMD on GPUs

From: Matthew Guberman-Pfeffer <matthew.guberman-pfeffer.uconn.edu>
Date: Thu, 4 Mar 2021 22:13:20 -0500

Dear Amber Community,

I am trying to run E-REMD on GPUs at an HPC. The job appears to be running from the perspective of the slurm queuing system, but when I check the activity on the GPUs, there are no running processes. I’m wondering if anyone might have a suggestion for settings I should check. My input files are given below.

Submission Script:
   #!/bin/bash

   #SBATCH -J prodWT
   #SBATCH -p gpu
   #SBATCH --gpus=8
   #SBATCH --ntasks=8
   #SBATCH --gpus-per-task=1
   #SBATCH --cpus-per-task=1
   #SBATCH -t 01:00:00

     module purge
     module load Amber/20-foss-2018b
     mpirun -np 8 pmemd.cuda_SPFP.MPI -ng 8 -groupfile groupfile

groupfile
# Replica 001
-O -rem 5 -remlog rem.log -remtype rem.info -i prod_1.rep.001 -p 2mz9-WT-WAT7937.new.prmtop -c ../cMD/equil_1.rst7 -x prod_1.mdcrd.001 -inf prod_1.mdinfo.001 -o prod_1.out.001 -r prod_1.rst7.001 -cpin 2mz9-WT-WAT7937.cpin -cpout 2mz9-WT-WAT7937.cpout.001 -cprestrt 2mz9-WT-WAT7937.cprestart.001 -cein 2mz9-WT-WAT7937.cein -ceout 2mz9-WT-WAT7937.ceout.001 -cerestrt 2mz9-WT-WAT7937.cerestart.001

# Replica 002



Best,
Matthew


 

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Mar 04 2021 - 19:30:02 PST
Custom Search