WhyTorch

WhyTorch is a website designed to explain PyTorch functions. Visit the live site.

Build and test locally

npm i
npx eleventy
DOCKER_BUILDKIT=0 docker buildx build . -t whytorch
docker stop whytorch
docker run --rm -d -p 80:80 -p 8080:8080 --name whytorch whytorch

Deployment notes

• WhyTorch is hosted using fly.io. It's accessible at both the root URL (see above), and on fly.io via https://whytorch.fly.dev.
• The PyTorch URL redirect is done automatically via the nginx configuration (nginx.conf).

Contributing

WhyTorch doesn't cover every function (try the sidebar on the site to see a list of unimplemented functions). That's where you can come in and help!

1. Find a function you want to document - it should already have a .js file in public/js/functions/
2. Read some existing files (public/js/functions/torch.add.js, public/js/functions/torch.gather.js)
3. Fork this repo
4. Write your function's js file
5. Add it to the list supportedMethods in public/js/main.js
6. Build and test locally using the instructions above
7. Commit, push, and merge request