Caffe

Description

Caffe is a deep learning framework made with expression, speed, and modularity in mind. It is developed by Berkeley AI Research (BAIR) and by community contributors.

Caffe supports many kinds of deep learning architectures geared towards image classification, segmentation, and speech recognition. It supports CNN, RCNN, LSTM, and fully connected neural network designs.

How to Build with SCALE

git clone https://github.com/BVLC/caffe.git
cd caffe
cmake \
    -DCMAKE_BUILD_TYPE=Release \
    -DUSE_CUDNN=OFF \
    -DCUDA_TOOLKIT_ROOT_DIR=/opt/scale \
    -DCUDA_ARCH_NAME=Manual \
    -DCUDA_ARCH_BIN="$GPU_ARCH" \
    -Bbuild .
make -C build -j$(nproc) install

Notes

• All standard layers with CUDA implementations are validated
• cuDNN integration is disabled in the current validation configuration
• Pre-trained model zoo is compatible