jqhoogland / arena

Repo for exercises from ARENA Virtual

Jesse's ARENA Virtual

Exercises for the 2022 ARENA Virtual Program.

Note This was forked from a template provided by Callum McDougall, who wrote most of the tests and utils. I've reorganized the repo to be more informative (and completed the exercises).

The exercises consist of a combination of:

• Recreating parts of torch to understand the internals (and build NNs from scratch).
• Reimplementing well-known models (AlexNet, ResNet, GPT-2, BERT, U-Net, etc.)