SourcesOriginal systems reading index and corpus software-layer organization; planned walkthrough exercises.
5.40.1Implementing a Mini Training Runtime from Scratch#
5.40.2Implementing a Continuous-Batching LLM Server from Scratch#
5.40.3The Complete Path of One Request Through SGLang/vLLM#
5.40.4The Complete Path of One RL Step Through Rollout/Reward/Training#
5.40.5From HF Checkpoint to Quantization, Multi-Device, and Disaggregated Deployment#
5.40.6From Model Frontend to Device Execution: Cross-Platform Paths Within Public Interfaces#
5.40.7Cross-Layer Diagnosis from Kernel to Compiler to Runtime to Driver to Hardware#
5.40.8Deployment Differences for the Same Model Across Hardware/Programming Models#