Validation Modules

trainModules checks the training primitives of the standard layout in isolation, on small deterministic inputs.

TrainingModules

trainModules runs single backward primitives of the standard controllers (ReLU, batch norm, fully connected, convolution, pooling, softmax and losses) and prints their results, to be compared with the independent NumPy computation of python/training/validate_modules.py.

Build and Run

cmake -DCMAKE_BUILD_TYPE=Release -DMODE=TRAINING ..
make -j$(nproc) trainModules
./trainModules

Full Example Source

You can view and download the full source code of this example:

TrainingModules.cpp

validate_modules.py