minggangzhao/shiftnet_cuda_v2

Shift Operation CUDA Implementation

created by Peter Jin

Tradeoffs and further analysis can be found in the paper. If you find this work useful for your research, please consider citing:

@inproceedings{shift,
    Author = {Bichen Wu and Alvin Wan and Xiangyu Yue and Peter Jin and Sicheng Zhao and Noah Golmant and Amir Gholaminejad and Joseph Gonzalez and Kurt Keutzer},
    Title = {Shift: A Zero FLOP, Zero Parameter Alternative to Spatial Convolutions},
    Journal = {arXiv:1711.08141},
    Year = {2017}
}

Uses of Shift:

• ShiftResNet for CIFAR-10, CIFAR-100 classification

Installation

If you have included this shift repository as a submodule in a separate repository, feel free to skip down to step 5.

1. If you have not already, setup a virtual environment with Python3, and activate it.

virtualenv shift --python=python3
source shift/bin/activate

Your prompt should now be prefaced with (shift), as in

(shift) [user@server:~]$ 

2. Install pytorch and torchvision. Access pytorch.org, scroll down to the "Getting Started" section, and select the appropriate OS, package manager, Python, and CUDA build. For example, selecting Linux, pip, Python3.5, and CUDA 8 gives the following, as of the time of this writing

pip3 install http://download.pytorch.org/whl/cu80/torch-0.3.0.post4-cp35-cp35m-linux_x86_64.whl 
pip3 install torchvision

3. Clone this repository.

git clone [email protected]:peterhj/shiftnet_cuda_v2.git

4. cd into the root of this repository.

cd shiftnet_cuda_v2

5. Install the Python requirements for this package.

pip3 install -r requirements.txt

6. Compile the Shift Layer implementation in C.

make

Getting invalid_device_function? Update the architecture code in models/shiftnet_cuda_v2/Makefile, currently configured for a Titan X. e.g., A Tesla K80 is sm-30.

Your custom CUDA layer is now installed.

Test

To check that the build completed successfully, run the test script

python test_shiftnet.py

After ~3s, the script should output a number of different tensors, where the last tensor has non-zero values only in the first column.

Columns 13 to 17
   89     0     0     0     0
  107     0     0     0     0
  125     0     0     0     0
  143     0     0     0     0
  161     0     0     0     0
  179     0     0     0     0
  197     0     0     0     0
  215     0     0     0     0
  233     0     0     0     0
  251     0     0     0     0
  269     0     0     0     0
  287     0     0     0     0
  305     0     0     0     0
  323     0     0     0     0
    0     0     0     0     0
    0     0     0     0     0
    0     0     0     0     0
    0     0     0     0     0
[torch.FloatTensor of size 18x18]