This is really cool, great way to reinforce our understanding of model architectures! But how is the author confirming that these model architecture implementations are correct though? I don't see any details in the README.md.