Frameworks and Models

The code a model is written in and the files it ships as: PyTorch, JAX, and Keras 3 side by side, then how a checkpoint on the Hugging Face Hub becomes weights on a GPU.

For: ML engineers onboarding to a codebase, and anyone who has to debug “the model will not load”.

Included in: Superstar FDE. Progress made here counts there too.

The stages, their chapters, and what “done” means for each are in path.tsv.

practice/bin/ol learn frameworks-and-models          # stages and progress
practice/bin/ol learn frameworks-and-models next     # read the next unfinished stage

Stages

Generated from path.tsv. Track progress locally with practice/bin/ol learn.

Stage	Read	Done when
1	Frameworks: PyTorch, JAX, Keras 3 Training & Frameworks	The same tiny GPT trains in PyTorch and in JAX, and you can explain eager vs traced execution and where torch.compile and jit differ.
2	Loading models from the Hub Model Loading: Hub to GPU	You inspect a checkpoint without downloading it (safetensors header, config, tokenizer, chat template) and can diagnose the five most common load failures.