🕶️ Classy Template

Welcome! If you're landing here, it's probably because you want to try out our library, classy.

As you might have read from our documentation, classy comes with a bunch of ready-to-use tools and tasks. But, for more particular settings (as yours is, probably!), we have developed this template that we have specifically built to seamlessly integrate with classy.

In case you didn't, check out classy's preconfigured tasks to make sure it doesn't cover your use case already! We cover quite a few tasks, including the most common ones! 🔥

If we did not match your needs (...sorry!), click on use this template at the top of this page to start.

💻 Creating your environment

The best way to start things off is by creating the environment. Once you have cloned your new repository locally, just type bash setup.sh and follow the setup script. It will automatically install all requirements listed inside requirements.txt along with classy (and its autocomplete).

And... You're all set! Yes, it's that easy.

❓ What can/should I do?

Well, there are a few (basic) things that you might be interested in doing:

• Creating a custom model (link)
• Creating a custom dataset (link)
• Creating a custom data reader (link)

📜 Examples

Additionally, that we have an ever-growing repository of examples available that touch most parts of classy's structure.

🤖 How does classy integrate with the template?

Great question!

For starters, classy will automatically discover any configuration files you add under configurations/ and will also automatically discover any 🐍 Python file defined under src/!

This way, you can easily change the model you train from the classy CLI by doing

classy train <task> <dataset> -n <exp-name> \
+      -c model._target_=src.model.MyCustomClassyModel

🆘 Help / get in touch

Finally, if you need help, want to report a bug, or just want to get in touch, open an issue on the main classy repository or join us on Slack!