test.sh

module rm compiler/rocm/2.9
source /public/home/aiss/Pytorch/rocm3.3_env_torch.sh
#for multinode
source `pwd`/config_DGX1_multinode.sh


SLURM_NTASKS_PER_NODE=${SLURM_NTASKS_PER_NODE:-$DGXNGPU}
SLURM_JOB_ID=${SLURM_JOB_ID:-$RANDOM}
echo "Run vars: id $SLURM_JOB_ID gpus $SLURM_NTASKS_PER_NODE mparams $MULTI_NODE"

set -e

# start timing
start=$(date +%s)
start_fmt=$(date +%Y-%m-%d\ %r)
echo "STARTING TIMING RUN AT $start_fmt"

# run benchmark
set -x
NUMEPOCHS=${NUMEPOCHS:-70}
LR=${LR:-"2.5e-3"}

echo "running benchmark"

export DATASET_DIR="/public/software/apps/DeepLearning/Data/COCO2017"

dist_url=`sed -n '1p' ./tmp`
#echo $dist_url

rank=0
num_lines=`cat ./tmp |wc -l`

for((i=0;i<$num_lines-1;i++))
do
   ((rank=$i+1))
    nodename=$(cat ./tmp |sed -n "${rank}p")
    ssh ${nodename} "module rm compiler/rocm/2.9 && source /public/home/aiss/Pytorch/rocm3.3_env_torch.sh && python3 -m bind_launch --nnodes=$num_lines --node_rank=$i  --master_addr=${dist_url} --master_port=4567 --nsockets_per_node ${DGXNSOCKET} --ncores_per_socket ${DGXSOCKETCORES} --nproc_per_node $SLURM_NTASKS_PER_NODE --no_hyperthreads --no_membind `pwd`/train.py --epochs ${NUMEPOCHS} --warmup-factor 0 --lr ${LR} --threshold=0.23 --data ${DATASET_DIR} ${EXTRA_PARAMS[@]}"
done

#((i=${num_lines}-1))
i=0
echo "lllllll"
nodename=$(cat ./tmp |sed -n "${num_lines}p")

ssh ${nodename} "cd `pwd` && module rm compiler/rocm/2.9 && source /public/home/aiss/Pytorch/rocm3.3_env_torch.sh && HIP_VISIBLE_DEVICES=0,1,2,3 python3 -m bind_launch  --nnodes=$num_lines --node_rank=$i  --master_addr=${dist_url} --master_port=4567 --nsockets_per_node ${DGXNSOCKET} --ncores_per_socket ${DGXSOCKETCORES} --nproc_per_node $SLURM_NTASKS_PER_NODE --no_hyperthreads --no_membind `pwd`/train.py --epochs ${NUMEPOCHS} --warmup-factor 0 --lr ${LR} --threshold=0.23 --data ${DATASET_DIR} ${EXTRA_PARAMS[@]} && ret_code=$?"
#ssh ${nodename} "cd `pwd` && module rm compiler/rocm/2.9 && source /public/home/aiss/Pytorch/rocm3.3_env_torch.sh && HIP_VISIBLE_DEVICES=0,1,2,3 python3 -m bind_launch  --nnodes=$num_lines --node_rank=$i  --master_addr=${dist_url} --master_port=4567 --nsockets_per_node ${DGXNSOCKET} --ncores_per_socket ${DGXSOCKETCORES} --nproc_per_node $SLURM_NTASKS_PER_NODE --no_hyperthreads --no_membind `pwd`/train.py --epochs ${NUMEPOCHS} --warmup-factor 0 --lr ${LR} --threshold=0.23 --data ${DATASET_DIR} --batch-size 64 --warmup 2.619685 && ret_code=$?"

#python3 -m bind_launch --nsockets_per_node ${DGXNSOCKET} \

#                      --ncores_per_socket ${DGXSOCKETCORES} \
#                      --nproc_per_node $SLURM_NTASKS_PER_NODE $MULTI_NODE \
#                      --no_hyperthreads \
#                      --no_membind \
# train.py \
#  --epochs "${NUMEPOCHS}" \
#  --warmup-factor 0 \
#  --lr "${LR}" \
#  --no-save \
#  --threshold=0.23 \
#  --data ${DATASET_DIR} \
#  ${EXTRA_PARAMS[@]} ; ret_code=$?

set +x

sleep 3
if [[ $ret_code != 0 ]]; then exit $ret_code; fi

# end timing
end=$(date +%s)
end_fmt=$(date +%Y-%m-%d\ %r)
echo "ENDING TIMING RUN AT $end_fmt"

# report result
result=$(( $end - $start ))
result_name="OBJECT_DETECTION"

echo "RESULT,$result_name,,$result,nvidia,$start_fmt"