Intel Deep Learning SDK

Vengineerの戯言 : Twitter
SystemVerilogの世界へようこそ、すべては、SystemC v0.9公開から始まった

4月9日のブログ、INTEL AND MLにも出てきた

をチェックしました。

Intel® Deep Learning SDKから
ツールは、「Training Tool」と「Deployment Tool」から構成される

引用
　・Training Tool

　　Easily prepare training data, design models, and train models with automated 
　　experiments and advanced visualizations
　　Simplify the installation and usage of popular deep learning frameworks 
　　optimized for Intel platforms

　　Data scientists are the primary users.

　・Deployment Tool

　　Optimize trained deep learning models through model compression and weight 
　　quantization, which are tailored to end-point device characteristics
　　Deliver a unified API to integrate the inference with application logic

　　Application developers are the primary users.

Training Toolは、データサイエンティスト用
Deployment Toolはアプリケーション開発者用

なので、ここからは、Deployment Toolのみを

引用
　　Required Hardware　　Optimized for Intel® Xeon® and Intel® Core™ processors
　　Required OS　　　　　Development Environment: Ubuntu 14.04, 16.04 (64 bit)
　　Target Inference Platform: 
　　　　　　　　　　　　Ubuntu 14.04 (64 bit)
　　Supported Use Cases	Image classification and image segmentation
　　Supported Layers	Convolution, deconvolution, fully connected, pooling, 
　　　　　　　　　　　　rectified linear unit (RelU), softmax, eltwise, crop, 
　　　　　　　　　　　　local response normalization (LRN), 
　　　　　　　　　　　　concatenation, power, and split

開発環境は、Ubuntu 14.04と16.04(64bit)のみ。
で、ターゲットもUbuntu 14.04(64bit)

えええええ、ということは組み込みには使えないということなのか？

ドキュメントもチェックしました。

Deplymento Toolは、下記の2つ(Model OptimizerとInference Engine)から構成される

引用
  Model Optimizer  

  Model Optimizer is a cross-platform command line tool that:
   • Takes as input a trained network that contains a certain network topology, 
     parameters, and the trained weights and biases. 
     The Model Optimizer currently only supports input networks 
     that are produced using the Caffe* framework.
   • Performs horizontal and vertical fusion of the network layers.
   • Prunes unused branches in the network.
   • Applies weights compression methods.
   • Produces as output an Internal Representation (IR) of the network 
     – a pair of files that describe the whole model:  
   • Topology file – an .xml file that describes the network topology. 
   • Trained data file 
     – a .bin file that contains the weights and biases as binary data. 
   • The produced IR is used as an input for the Inference Engine. 

  Inference Engine  

  Inference Engine is a runtime which: 
   • Takes as input an IR produced by Model Optimizer 
   • Optimizes inference execution for target hardware 
   • Delivers inference solution with reduced footprint on embedded inference platforms
   • Enables seamless integration with application logic, 
     which eases transition between platforms from Intel® through 
     supporting the same API across a variety of platforms.

Model Optimizerが.prototxtと.caffemodelを読み込み、IR(.xmlと.bin)を出力する
そのIRをInference Engineを組み込んだアプリケーションが読み込むと。。。

そして、Model Optimizerの制約は。。。

引用
　　• It is distributed for 64-bit Ubuntu* OS 14.04 only.
　　• It can process models in Caffe* format only.
　　• It can process popular image classification network models, 
      including AlexNet, GoogleNet, VGG-16, LeNet, and ResNet-152, 
      and fully convolutional network models like FCN8 
      that are used for image segmentation. 
      It may fail to support a custom network.

It may fail to support a custom network

え、何？