Started by timer
Running as SYSTEM
Building in workspace /var/lib/jenkins/jobs/pytorch_infer/workspace
[SSH] script:
TARGETNODE=""""
module load anaconda3_gpu/4.13.0
module load cuda/11.7.0
cd pytorch_infer
rm -f infer_results_jenkins.csv
# Slurm Arguments
sargs="--nodes=1 "
sargs+="--ntasks-per-node=1 "
sargs+="--mem=16g "
sargs+="--time=00:10:00 "
sargs+="--account=bbmb-hydro "
sargs+="--gpus-per-node=1 "
sargs+="--gpu-bind=closest "
# Add Target node if it exists
if [[ ! -z ${TARGETNODE} ]]
then
PARTITION=`sinfo --format="%R,%N" -n hydro61 | grep hydro61 | cut -d',' -f1 | tail -1`
sargs+="--partition=${PARTITION} "
sargs+="--nodelist=${TARGETNODE} "
else
sargs+="--partition=a100 "
fi
# Executable to run
scmd="python benchmark.py --model-list jenkins_list_short.txt --bench inference --channels-last --results-file infer_results_jenkins.csv"
# Run the command
start_time=`date +%s.%N`
echo $"Starting srun with command"
echo "srun $sargs $scmd"
srun $sargs $scmd
end_time=`date +%s.%N`
python transpose_results.py
runtime=$( echo "$end_time - $start_time" | bc -l )
echo "YVALUE=$runtime" > time.txt
printf "Pytorch test completed in %0.3f secs\n" $runtime
[SSH] executing...
Starting srun with command
srun --nodes=1 --ntasks-per-node=1 --mem=16g --time=00:10:00 --account=bbmb-hydro --gpus-per-node=1 --gpu-bind=closest --partition=a100 python benchmark.py --model-list jenkins_list_short.txt --bench inference --channels-last --results-file infer_results_jenkins.csv
srun: job 98789 queued and waiting for resources
srun: job 98789 has been allocated resources
Running benchmark on hydro03
Running bulk validation on these pretrained models: vgg19_bn, resnet18, resnet34, simplenetv1_5m_m1,
Benchmarking in float32 precision. NHWC layout. torchscript disabled
Model vgg19_bn created, param count: 143678248
Running inference benchmark on vgg19_bn for 40 steps w/ input size (3, 224, 224) and batch size 256.
Infer [8/40]. 1429.67 samples/sec. 179.063 ms/step.
Infer [16/40]. 1430.13 samples/sec. 179.004 ms/step.
Infer [24/40]. 1429.48 samples/sec. 179.086 ms/step.
Infer [32/40]. 1429.31 samples/sec. 179.108 ms/step.
Infer [40/40]. 1429.24 samples/sec. 179.116 ms/step.
Inference benchmark of vgg19_bn done. 1429.04 samples/sec, 179.12 ms/step
Benchmarking in float32 precision. NHWC layout. torchscript disabled
Model resnet18 created, param count: 11689512
Running inference benchmark on resnet18 for 40 steps w/ input size (3, 224, 224) and batch size 256.
Infer [8/40]. 10657.94 samples/sec. 24.020 ms/step.
Infer [16/40]. 10662.23 samples/sec. 24.010 ms/step.
Infer [24/40]. 10649.65 samples/sec. 24.038 ms/step.
Infer [32/40]. 10618.20 samples/sec. 24.110 ms/step.
Infer [40/40]. 10620.10 samples/sec. 24.105 ms/step.
Inference benchmark of resnet18 done. 10615.67 samples/sec, 24.11 ms/step
Benchmarking in float32 precision. NHWC layout. torchscript disabled
Model resnet34 created, param count: 21797672
Running inference benchmark on resnet34 for 40 steps w/ input size (3, 224, 224) and batch size 256.
Infer [8/40]. 6492.84 samples/sec. 39.428 ms/step.
Infer [16/40]. 6468.03 samples/sec. 39.579 ms/step.
Infer [24/40]. 6467.54 samples/sec. 39.582 ms/step.
Infer [32/40]. 6472.98 samples/sec. 39.549 ms/step.
Infer [40/40]. 6472.86 samples/sec. 39.550 ms/step.
Inference benchmark of resnet34 done. 6471.25 samples/sec, 39.55 ms/step
Benchmarking in float32 precision. NHWC layout. torchscript disabled
Model simplenetv1_5m_m1 created, param count: 5752808
Running inference benchmark on simplenetv1_5m_m1 for 40 steps w/ input size (3, 224, 224) and batch size 256.
Infer [8/40]. 13988.36 samples/sec. 18.301 ms/step.
Infer [16/40]. 13985.35 samples/sec. 18.305 ms/step.
Infer [24/40]. 13983.98 samples/sec. 18.307 ms/step.
Infer [32/40]. 13987.82 samples/sec. 18.302 ms/step.
Infer [40/40]. 13980.69 samples/sec. 18.311 ms/step.
Inference benchmark of simplenetv1_5m_m1 done. 13973.98 samples/sec, 18.31 ms/step
args: Namespace(model_list='jenkins_list_short.txt', bench='inference', detail=False, results_file='infer_results_jenkins.csv', num_warm_iter=10, num_bench_iter=40, model='vgg19_bn', batch_size=256, img_size=None, input_size=None, use_train_size=False, num_classes=None, gp=None, channels_last=True, grad_checkpointing=False, amp=False, precision='float32', torchscript=False, fuser='', opt='sgd', opt_eps=None, opt_betas=None, momentum=0.9, weight_decay=0.0001, clip_grad=None, clip_mode='norm', smoothing=0.1, drop=0.0, drop_path=None, drop_block=None)
args: Namespace(model_list='jenkins_list_short.txt', bench='inference', detail=False, results_file='infer_results_jenkins.csv', num_warm_iter=10, num_bench_iter=40, model='resnet18', batch_size=256, img_size=None, input_size=None, use_train_size=False, num_classes=None, gp=None, channels_last=True, grad_checkpointing=False, amp=False, precision='float32', torchscript=False, fuser='', opt='sgd', opt_eps=None, opt_betas=None, momentum=0.9, weight_decay=0.0001, clip_grad=None, clip_mode='norm', smoothing=0.1, drop=0.0, drop_path=None, drop_block=None)
args: Namespace(model_list='jenkins_list_short.txt', bench='inference', detail=False, results_file='infer_results_jenkins.csv', num_warm_iter=10, num_bench_iter=40, model='resnet34', batch_size=256, img_size=None, input_size=None, use_train_size=False, num_classes=None, gp=None, channels_last=True, grad_checkpointing=False, amp=False, precision='float32', torchscript=False, fuser='', opt='sgd', opt_eps=None, opt_betas=None, momentum=0.9, weight_decay=0.0001, clip_grad=None, clip_mode='norm', smoothing=0.1, drop=0.0, drop_path=None, drop_block=None)
args: Namespace(model_list='jenkins_list_short.txt', bench='inference', detail=False, results_file='infer_results_jenkins.csv', num_warm_iter=10, num_bench_iter=40, model='simplenetv1_5m_m1', batch_size=256, img_size=None, input_size=None, use_train_size=False, num_classes=None, gp=None, channels_last=True, grad_checkpointing=False, amp=False, precision='float32', torchscript=False, fuser='', opt='sgd', opt_eps=None, opt_betas=None, momentum=0.9, weight_decay=0.0001, clip_grad=None, clip_mode='norm', smoothing=0.1, drop=0.0, drop_path=None, drop_block=None)
--result
[
{
"model": "simplenetv1_5m_m1",
"infer_samples_per_sec": 13973.98,
"infer_step_time": 18.311,
"infer_batch_size": 256,
"infer_img_size": 224,
"param_count": 5.75
},
{
"model": "resnet18",
"infer_samples_per_sec": 10615.67,
"infer_step_time": 24.105,
"infer_batch_size": 256,
"infer_img_size": 224,
"param_count": 11.69
},
{
"model": "resnet34",
"infer_samples_per_sec": 6471.25,
"infer_step_time": 39.55,
"infer_batch_size": 256,
"infer_img_size": 224,
"param_count": 21.8
},
{
"model": "vgg19_bn",
"infer_samples_per_sec": 1429.04,
"infer_step_time": 179.116,
"infer_batch_size": 256,
"infer_img_size": 224,
"param_count": 143.68
}
]
Pytorch test completed in 62.550 secs
[SSH] completed
[SSH] exit-status: 0
[workspace] $ /bin/sh -xe /tmp/jenkins5730827684215187114.sh
+ scp 'HYDRO_REMOTE:~svchydrojenkins/pytorch_infer/time.txt' /var/lib/jenkins/jobs/pytorch_infer/workspace
+ scp 'HYDRO_REMOTE:~svchydrojenkins/pytorch_infer/infer_results_jenkins.csv' /var/lib/jenkins/jobs/pytorch_infer/workspace
Recording plot data
Saving plot series data from: /var/lib/jenkins/jobs/pytorch_infer/workspace/time.txt
Finished: SUCCESS
![[Jenkins]](/static/755cd4af/images/svgs/logo.svg){kind=logo}
