Add new job file for running horovod benchmarks

horovod-benchmark.job

+#!/bin/bash
+#SBATCH --share
+#SBATCH --partition=pascalnodes
+#SBATCH --exclusive
+#
+# Name your job to make it easier for you to track
+#
+#SBATCH --job-name=keras_mpi
+#
+# Set your error and output files
+#
+#SBATCH --error=keras_resnet_3node_mpi.err
+#SBATCH --output=keras_resnet_3node_mpi.out
+#SBATCH --ntasks=12
+#SBATCH --gres=gpu:4
+#SBATCH -N3
+#
+# Tell the scheduler only need 12 hours
+#
+#SBATCH --time=12:00:00
+#SBATCH --cpus-per-task=2
+#SBATCH --mem-per-cpu=16000
+#
+# Set your email address and request notification when you job is complete or if it fails
+#
+#SBATCH --mail-type=FAIL
+#SBATCH --mail-user=$USER@uab.edu
+
+module load Anaconda3/5.2.0
+
+module load cuda91/toolkit
+
+module load OpenMPI/3.1.2-gcccuda-2018b
+
+source activate distributedLearning
+
+time mpirun -np $SLURM_NTASKS -bind-to none -map-by slot -mca pml ob1 -mca btl_tcp_if_include ib0 python /data/user/$USER/benchmarks/scripts/tf_cnn_benchmarks/tf_cnn_benchmarks.py --model resnet101 --batch_size 64 --variable_update horovod
\ No newline at end of file