3 [![Code Coverage](http://ec2-54-180-96-14.ap-northeast-2.compute.amazonaws.com/nntrainer/ci/badge/codecoverage.svg)](http://ec2-54-180-96-14.ap-northeast-2.compute.amazonaws.com/nntrainer/ci/gcov_html/index.html)
4 ![GitHub repo size](https://img.shields.io/github/repo-size/nnstreamer/nntrainer)
5 ![GitHub issues](https://img.shields.io/github/issues/nnstreamer/nntrainer)
6 ![GitHub pull requests](https://img.shields.io/github/issues-pr/nnstreamer/nntrainer)
8 NNtrainer is Software Framework for Training Neural Network Models on Devices.
12 NNtrainer is an Open Source Project. The aim of the NNtrainer is to develop Software Framework to train neural network model on embedded devices which has relatively limited resources. Rather than training the whole layers, NNtrainer trains only one or a few layers added after the feature extractor.
14 Even though it trains part of the neural network models, NNtrainer requires quite a lot of functionalities to train from common neural network frameworks. By implementing them, it is good enough to run several examples which can help to understand how it works. There are KNN, Neural Network, Logistic Regression and Reinforcement Learning with CartPole in Applications directory and some of them use Mobilenet V2 with tensorflow-lite as a feature extractor. All of them tested on Galaxy S8 with Android and PC (Ubuntu 16.04).
17 * [Jijoong Moon](https://github.com/jijoongmoon)
18 * [MyungJoo Ham](https://github.com/myungjoo)
19 * [Geunsik Lim](https://github.com/leemgs)
22 * [Sangjung Woo](https://github.com/again4you)
23 * [Wook Song](https://github.com/wooksong)
24 * [Jaeyun Jung](https://github.com/jaeyun-jung)
25 * [Hyoungjoo Ahn](https://github.com/helloahn)
26 * [Parichay Kapoor](https://github.com/kparichay)
27 * [Dongju Chae](https://github.com/dongju-chae)
28 * [Gichan Jang](https://github.com/gichan-jang)
29 * [Yongjoo Ahn](https://github.com/anyj0527)
35 This is the component which controls neural network layers. Read the configuration file ([Iniparser](https://github.com/ndevilla/iniparser) is used to parse the configuration file.) and constructs Layers including Input and Output Layer, according to configured information by the user.
36 The most important role of this component is to activate forward / backward propagation. It activates inferencing and training of each layer while handling the data properly among them. There are properties to describe network model as below:
38 - **_Type:_** Network Type - Regression, KNN, NeuralNetwork
39 - **_Layers:_** Name of Layers of Network Model
40 - **_Learning\_rate:_** Learning rate which is used for all Layers
41 - **_Decay\_rate:_** Rate for Exponentially Decayed Learning Rate
42 - **_Epoch:_** Max Number of Training Iteration.
43 - **_Optimizer:_** Optimizer for the Network Model - sgd, adam
44 - **_Activation:_** Activation Function - sigmoid , tanh
45 - **_Cost:_** Cost Function -
46 msr(mean square root error), categorical (for logistic regression), cross (cross entropy)
47 - **_Model:_** Name of Model. Weight Data is saved in the name of this.
48 - **_minibach:_** mini batch size
49 - **_beta1,beta2,epsilon:_** hyper parameters for the adam optimizer
54 This component defines Layers which consist of Neural Network Model. Every neural network model must have one Input & Output Layer and other layers such as Fully Connected or Convolution can be added between them. (For now, supports Input & Output Layer, Fully Connected Layer.)
56 - **_Type:_** InputLayer, OutputLayer, FullyConnectedLayer
57 - **_Id:_** Index of Layer
58 - **_Height:_** Height of Weight Data (Input Dimension)
59 - **_Width:_** Width of Weight Data ( Hidden Layer Dimension)
60 - **_Bias\_zero:_** Boolean for Enable/Disable Bias
65 Tensor is responsible for the calculation of Layer. It executes the addition, division, multiplication, dot production, averaging of Data and so on. In order to accelerate the calculation speed, CBLAS (C-Basic Linear Algebra: CPU) and CUBLAS (CUDA: Basic Linear Algebra) for PC (Especially NVIDIA GPU) for some of the operation. Later, these calculations will be optimized.
71 The following dependencies are needed to compile / build / run.
75 * blas library ( CBLAS ) (for CPU Acceleration)
76 * cuda, cudart, cublas (should match the version) (GPU Acceleration on PC)
77 * tensorflow-lite (>=1.4.0)
78 * jsoncpp ( >=0.6.0) (openAI Environment on PC)
79 * libcurl3 (>= 7.47 ) (openAI Environment on PC)
83 Download the source file by cloning the github repository.
86 $ git clone https://github.com/nnstreamer/nntrainer
89 After completing download the sources, you can find the several directories and files as below.
110 f1a3a05 (HEAD -> master, origin/master, origin/HEAD) Add more badges
111 37032a1 Add Unit Test Cases for Neural Network Initialization
112 181a003 lower case for layer type.
113 1eb399b Update clang-format
114 87f1de7 Add Unit Test for Neural Network
115 cd5c36e Add Coverage Test badge for nntrainer
119 You can find the source code of the core library in nntrainer/src. In order to build them, use [meson](https://mesonbuild.com/)
122 The Meson build system
124 Source dir: /home/wook/Work/NNS/nntrainer
125 Build dir: /home/wook/Work/NNS/nntrainer/build
126 Build type: native build
127 Project name: nntrainer
128 Project version: 0.0.1
129 Native C compiler: cc (gcc 7.5.0 "cc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0")
130 Native C++ compiler: c++ (gcc 7.5.0 "c++ (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0")
131 Build machine cpu family: x86_64
132 Build machine cpu: x86_64
134 Build targets in project: 11
135 Found ninja-1.8.2 at /usr/bin/ninja
138 ninja: Entering directory `build'
139 [41/41] Linking target test/unittest/unittest_nntrainer_internal.
142 After completion of the build, the shared library, 'libnntrainer.so' and the static library, 'libnntrainer.a' will be placed in build/nntrainer.
144 $ ls build/nntrainer -1
145 d48ed23@@nntrainer@sha
146 d48ed23@@nntrainer@sta
151 In order to install them with related header files to your system, use the 'install' sub-command.
153 $ ninja -C build install
155 Then, you will find the libnntrainer.so and related .h files in /usr/local/lib and /usr/local/include directories.
157 By default, the command ```ninja -C build`` generates the five example application binaries (Classification, KNN, LogisticRegression, ReinforcementLearning, and Training) you could try in build/Applications. For 'Training' as an example case,
159 $ ls build/Applications/Training/jni/ -1
160 e189c96@@nntrainer_training@exe
164 In order to run such example binaries, Tensorflow-lite is a prerequisite. If you are trying to run on the Android, it will automatically download tensorflow (1.9.0) and compile as static library. Otherwise, you need to install it by yourself.
166 To run the 'Training' example, do as follows
170 $ LD_LIBRARY_PATH=./build/nntrainer ./build/Applications/Training/jni/nntrainer_training ./Applications/Training/res/Training.ini ./Applications/Training/res/
171 ../../res/happy/happy1.bmp
172 ../../res/happy/happy2.bmp
173 ../../res/happy/happy3.bmp
174 ../../res/happy/happy4.bmp
175 ../../res/happy/happy5.bmp
176 ../../res/sad/sad1.bmp
177 ../../res/sad/sad2.bmp
183 ## Open Source License
185 The nntrainer is an open source project released under the terms of the Apache License version 2.0.