N3: Neural Network Notation

This project is WIP. Please be aware of using it.

node LeNet5:
    let K: kernel size = int 5

    let W: width = int 28
    let H: height = int 28

    with Conv2D:
        set kernel size = K
        set padding = K / 2
        set stride = 2

    node MyConv:

        1. Conv2D
        2. Relu

    0. Input                   =  1, W  , H
    1. MyConv                  = 32, W/2, H/2
    2. MyConv                  = 64, W/4, H/4
    3. Transform               = 64* W/4* H/4
    4. Linear + Relu + Dropout = 1024
    5. Linear + Softmax(D=-1)  = 10

Usage

Server

$ sudo systemctl start n3-torchd

Client

Training

$ n3 train image_classification --model LeNet5 --data MNIST --devices cuda:0 cpu

Evaluating

$ n3 eval image_classification --model LeNet5 --data MNIST --devices cuda:0 cpu

Publishing

$ n3 publish image_classification --model LeNet5 --target android:java

• android: java, flutter
• ios: flutter
• universal: c++, python

Monitoring using Tensorboard

$ n3 monitor # or, browse http://localhost::xxxx/

Distributed Training

$ n3 train image_classification --model LeNet5 --data MNIST --devices w:180:cuda:0 w:192.168.0.181 cpu

• "w:180:cuda:0": the "cuda:0" device in "xxx.xxx.xxx.180" (local)
• "w:192.168.0.181": automatically choose devices in "192.168.0.181"
• These can be defined as environment variables (N3_MACHINES)