Free Documentation

Compute and compilers

• CUDA C++ Programming Guide — NVIDIA execution and memory semantics.
• HIP Documentation — AMD heterogeneous-compute concepts and APIs.
• Triton Language Tutorials — tiled kernel exercises and code.
• Apache TVM Documentation — scheduling, lowering, and compiler workflows.

Distributed execution

• PyTorch Distributed Overview — framework APIs, process groups, and training patterns.
• NVIDIA Collective Communications Library Documentation — GPU collective semantics and topology.

Serving and evaluation

• vLLM Documentation — cache-aware, high-throughput LLM serving.
• NVIDIA Triton Inference Server Documentation — production model-serving controls.
• Holistic Evaluation of Language Models — scenario-based, multi-metric evaluation.
• Language Model Evaluation Harness — reproducible task and metric execution.
• PyTorch Tutorials — framework-level experiments across modules.

Documentation versions move. Record what was evaluated and when, and use Module 07 for an ordered distributed-systems path.