Validated workloads by framework

We validated performance [1] for the following TensorFlow* and MXNet* workloads:

TensorFlow

TensorFlow Workload Genre of Deep Learning
Resnet50 v1 Image recognition
Resnet50 v2 Image recognition
Inception V3 Image recognition
Inception V4 Image recognition
Inception-ResNetv2 Image recognition
MobileNet v1 Image recognition
MobileNet v2 Image recognition
VGG16 Image recognition
SSD-VGG16 Object detection
SSD-MobileNetv1 Object detection
R-FCN Object detection
Faster RCNN Object detection
Yolo v2 Object detection
Transformer-LT Language translation
Wide & Deep Recommender system
NCF Recommender system
U-Net Image segmentation
DCGAN Generative adversarial network
DRAW Image generation
A3C Reinforcement learning

MXNet

MXNet Workload Genre of Deep Learning
Resnet50 v1 Image recognition
Resnet50 v2 Image recognition
DenseNet-121 Image recognition
InceptionV3 Image recognition
InceptionV4 Image recognition
Inception-ResNetv2 Image recognition
MobileNet v1 Image recognition
SqueezeNet v1 and v1.1 Image recognition
VGG16 Image recognition
Faster RCNN Object detection
SSD-VGG16 Object detection
GNMT Language translation
Transformer-LT Language translation
Wide & Deep Recommender system
WaveNet Speech generation
DeepSpeech2 Speech recognition
DCGAN Generative adversarial network
A3C Reinforcement learning

ONNX

Additionally, we validated the following workloads are functional through nGraph ONNX importer. ONNX models can be downloaded from the ONNX Model Zoo.

ONNX Workload Genre of Deep Learning
ResNet-50 Image recognition
ResNet-50-v2 Image recognition
DenseNet-121 Image recognition
Inception-v1 Image recognition
Inception-v2 Image recognition
Mobilenet Image recognition
Shufflenet Image recognition
SqueezeNet Image recognition
VGG-19 Image recognition
ZFNet-512 Image recognition
MNIST Image recognition
Emotion-FERPlus Image recognition
BVLC AlexNet Image recognition
BVLC GoogleNet Image recognition
BVLC CaffeNet Image recognition
BVLC R-CNN ILSVRC13 Object detection
ArcFace Face Detection and Recognition

Important

Please see Intel’s Optimization Notice for details on disclaimers.

Footnotes

[1]Benchmarking performance of DL systems is a young discipline; it is a good idea to be vigilant for results based on atypical distortions in the configuration parameters. Every topology is different, and performance changes can be attributed to multiple causes. Also watch out for the word “theoretical” in comparisons; actual performance should not be compared to theoretical performance.