Skip to content

Latest commit

 

History

History
10 lines (5 loc) · 723 Bytes

README.md

File metadata and controls

10 lines (5 loc) · 723 Bytes

Implemented convolution based on CUDA extensions in PyTorch

A convolution implementation based on cuda extension for PyTorch. The source code reference to the PyTorch's inefficient implementation here. See here for the accompanying tutorial.

  • Build CUDA extensions by going into the conv_cuda/ folder and executing python setup.py install,
  • JIT-compile CUDA extensions by going into the conv_cuda/ folder and calling python jit.py, which will JIT-compile the extension and load it,
  • Check the result of the convolution by running python test.py