https://github.com/frgfm/torch-cuda-template

Template for CUDA / C++ extension writing with PyTorch
https://github.com/frgfm/torch-cuda-template

cpp cuda pytorch pytorch-extension

Last synced: 2 months ago
JSON representation

Template for CUDA / C++ extension writing with PyTorch

• Host: GitHub
• URL: https://github.com/frgfm/torch-cuda-template
• Owner: frgfm
• License: mit
• Created: 2020-09-08T20:47:59.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-09-09T12:55:44.000Z (over 4 years ago)
• Last Synced: 2024-11-22T13:07:11.303Z (3 months ago)
• Topics: cpp, cuda, pytorch, pytorch-extension
• Language: Python
• Homepage:
• Size: 14.6 KB
• Stars: 6
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • Contributing: CONTRIBUTING.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# torch-cuda-template

[![License](https://img.shields.io/badge/License-MIT-brightgreen.svg)](LICENSE) [![Codacy Badge](https://app.codacy.com/project/badge/Grade/9a56f466d21f4dbf8de677fd5b8709d3)](https://www.codacy.com/manual/frgfm/torch-cuda-template?utm_source=github.com&utm_medium=referral&utm_content=frgfm/torch-cuda-template&utm_campaign=Badge_Grade)![Build Status](https://github.com/frgfm/torch-cuda-template/workflows/python-package/badge.svg) [![codecov](https://codecov.io/gh/frgfm/torch-cda-template/branch/master/graph/badge.svg)](https://codecov.io/gh/frgfm/torch-cuda-template) 

Template for CUDA / C++ extension writing with PyTorch

## Table of Contents

- [Getting Started](#getting-started)

  - [Prerequisites](#prerequisites)

  - [Installation](#installation)

- [Usage](#usage)

- [Contributing](#contributing)

- [Credits](#credits)

- [License](#license)

## Getting started

### Prerequisites

- Python 3.6 (or more recent)

- [pip](https://pip.pypa.io/en/stable/)

### Installation

You can install the package using the default method:

```bash

pip install -e . --upgrade

```

which will build the CUDA extension and bind it with your python module.

## Usage

This python package can be used like any other:

```python

import torch

import cuda_ext.nn as nn

my_module = nn.DSigmoid()

with torch.no_grad():

	out = my_module(torch.rand(2, 3, 32, 32).cuda())

```

## Contributing

Please refer to `CONTRIBUTING` if you wish to contribute to this project.

## Credits

The content of this repo was brought together by the repo owner but highly benefited from the following resources:

- CUDA extensions of [Thomas Brandon](https://github.com/thomasbrandon)

- Official PyTorch [documentation](https://pytorch.org/tutorials/advanced/cpp_extension.html)

## License

Distributed under the MIT License. See `LICENSE` for more information.