KTransformers

GitHub Docs Reading Map

Early KTransformers GitHub docs contain current paths, historical entry points, and design background. Use this page to map older material to the current website.

Current User Docs

GitHub topicCurrent page
KT-Kernel install and SGLang integrationInstallation, Inference Overview
kt CLICLI Reference
MiniMax M2/M2.1, Kimi K2 Thinking, DeepSeek V4-FlashModel-specific pages under Supported Models
LLaMA-Factory KT SFT quick startFine-Tuning

Technical Background

GitHub topicCurrent page
AMX architecture notesAMX Backend and this Technical Work section
Native precision and conversion toolsPrecision and Quantization
Expert placement and schedulingExpert Placement
Public PDFs and talksTalks and Slides

Keep as Historical

Pages based on local_chat.py, ktransformers/server/main.py, balance_serve, old kt-sft, kt_optimize_rule, old NPU/xPU paths, or old Docker flows are historical material. Before using them, check whether a current package path and smoke result exist.