Kata Containers provides a series of performance tests using the TensorFlow reference benchmarks (tf_cnn_benchmarks). The tf_cnn_benchmarks containers TensorFlow implementations of several popular convolutional models https://github.com/tensorflow/benchmarks/tree/master/scripts/tf_cnn_benchmarks.
Currently the TensorFlow benchmark on Kata Containers includes test for
the AxelNet
and ResNet50
models.
Individual tests can be run by hand, for example:
$ cd metrics/machine_learning
$ ./tensorflow_nhwc.sh 25 60
Based on a suite of Python high performance computing benchmarks that uses various popular Python HPC libraries using Python https://github.com/dionhaefner/pyhpc-benchmarks.
Individual tests can be run by hand, for example:
$ cd metrics/machine_learning
$ ./pytorch.sh 40 100
MobileNets
are small, low-latency, low-power models parameterized to meet the resource
constraints of a variety of use cases. They can be built upon for classification, detection,
embeddings and segmentation similar to how other popular large scale models, such as Inception, are used.
MobileNets
can be run efficiently on mobile devices with Tensorflow
Lite.
Kata Containers provides a test for running MobileNet V1
inference using Intel-Optimized TensorFlow
.
Individual test can be run by hand, for example:
$ cd metrics/machine_learning
$ ./tensorflow_mobilenet_benchmark.sh 25 60
ResNet50
is an image classification model pre-trained on the ImageNet
dataset.
Kata Containers provides a test for running ResNet50
inference using Intel-Optimized
TensorFlow
.
Individual test can be run by hand, for example:
$ cd metrics/machine_learning
$ ./tensorflow_resnet50_int8.sh 25 60