Project documentation beyond the root README.
Detailed flowcharts of two quantization processes:
- Offline quantization — what happens when you run
nanoptq quantize(step-by-step, with math) - Runtime inference — what happens in
QuantLinear.forward()at serving time
Read these to understand the full end-to-end lifecycle of a quantized model.
Every quantization term explained with an analogy first, then a formal definition. Covers: quantization, scale, zero-point, group-wise, perplexity, RTN, AWQ, GPTQ, calibration data, int4/int8/fp8/bf16/nvfp4, Hessian, safetensors, outlier channels.
Start here if any term in the codebase is unfamiliar.
Internal planning and development tooling (AI agent skill files and plan documents). Not relevant to learning quantization.