developer_guide.md

We recommend you read through this document before you develope something new.

• [Use test]( #Use test )
• [Write new test cases](#Write new test case)

Use test

We provide some cases to validate our cpp modules.

Build test

During 'Install torch wrapper' step, add -DBUILD_TESTS=ON to cmake setting.

For CPU version:

cmake -DBUILD_TESTS=ON -DCMAKE_INSTALL_PREFIX=/usr/local/cpp-torch -DCMAKE_PREFIX_PATH=/usr/local/cpp-torch ..

For GPU version:

cmake -DBUILD_TESTS=ON -DBUILD_CUDA=ON -DCMAKE_INSTALL_PREFIX=/usr/local/cpp-torch -DCMAKE_PREFIX_PATH=/usr/local/cpp-torch ..

Run test

Go to cpp-torch/build/ and run 'ctest' in command. You will get something like this:

92% tests passed, 2 tests failed out of 24

Total Test time (real) =  11.16 sec

The following tests FAILED:
          7 - cpptorch_ConvolutionLayers_Spatial_SpatialBatchNormalization_2.lua (Failed)
         14 - cpptorch_dpnn_Decorator_1.lua (Failed)

Write new test case

Organization

Each .lua file under test/cases is a test case. Please follow torch's nn package document to organize cases as follows:

├─ Containers
├─ ConvolutionLayers
│  ├─ Spatial
│  ├─ Temporal
│  └─ Volumetric
├─ dpnn
├─ SimpleLayers
│  ├─ BasicTensor
│  ├─ MathTensor
│  ├─ Miscellaneous
│  └─ Parameterized
├─ TableLayers
│  ├─ CMath
│  ├─ Container
│  ├─ Conversion
│  ├─ Criteria
│  └─ Pair
└─ TransferFunctions

A single module can have multiple test cases, e.g.: Linear_1.lua, Linear_2.lua...

Write a test case

Check out this test cases for nn.RELU:

require 'torch'
require 'nn'
torch.setdefaulttensortype('torch.FloatTensor')
-- create input, output and net
input = torch.Tensor(6,3,128,128)
net = nn.ReLU()
net:evaluate()
output = net:forward(input)
-- save
torch.save(arg[1], input)
torch.save(arg[2], output)
net:clearState()
torch.save(arg[3], net)

Test case saves input, output and net to destination denoted by the 3 command line args.

Remember to re-build cpp-torch after you add new test cases.

So what's happening?

When you run ctest under /build, it performs the following actions:

• run each .lua file in torch, save input tensor, output tensor and net as .t7 file.
• feed input tensor into net in cpp-torch, compare output tensor with saved output tensor.

Write torch layer(TODO)

Write self-defined layer(TODO)