Skip to content

INSTALL.md

Latest commit

 

History

History
57 lines (44 loc) · 1.76 KB

INSTALL.md

File metadata and controls

57 lines (44 loc) · 1.76 KB

TASO Installation

TASO can be built from source code using the following instructions. We also provide prebuilt TASO docker images with all dependencies preinstalled.

Install from Source

Prerequisties

  • CMAKE 3.2 or higher
  • ProtocolBuffer 3.6.1 or higher
  • Cython 0.28 or higher
  • ONNX 1.5 or higher
  • CUDA 9.0 or higher and CUDNN 7.0 or higher

Build TASO Runtime

  • To get started, clone the TASO source code from github.
git clone --recursive https://www.github.com/jiazhihao/taso
cd taso

The TASO_HOME environment is used for building and running TASO. You can add the following line in ~/.bashrc.

export TASO_HOME=/path/to/taso
  • Build the TASO runtime library. The configuration of the TASO runtime can be modified by config.cmake. The default configuration builds the CUDA backend and automatically finds the CUDA libraries (e.g., cuDNN, cuBLAS). You can manually choose a CUDA path by changing set(USE_CUDA ON) to set(USE_CUDA /path/to/cuda/library). MKL support is coming soon.
mkdir build; cd build; cmake ..
sudo make install -j 4
  • Install the TASO python package.
cd ../python
python setup.py install

Docker Images

We require docker and nvidia-docker to run the TASO docker images.

  • First, clone the TASO gitpub repository to obtain the necessary scripts
git clone --recursive https://www.github.com/jiazhihao/taso
  • Second, we can use the following command to run a TASO docker image for CUDA 10.0.
/path/to/taso/docker/run_docker.sh tasoml/cuda100
  • You are ready to use TASO now. Try some of our example DNN architectures.
python /path/to/taso/examples/resnext10.py