Foundations
The core ML concepts every practitioner should be able to write code for: backprop, the training loop, transformers, and PyTorch fluency.
What’s here
- backprop/ — what backpropagation is, why it works, hand-rolling it.
- transformers/ — attention, the transformer block, positional encodings, layer norm.
- training-loop/ — the standard PyTorch training loop, optimizers, schedulers.
- pytorch/ — getting fluent in PyTorch as a tool.
Status
[~] In progress.
Prerequisites
prereqs/ — at least linear algebra basics. Backprop needs the chain rule.
Resources
- (placeholder — add specific tutorials, blog posts, and lectures as I work through them)