Writing

Writing

Engineering records, mechanisms and research notes.

Blog & notes

All posts →Write

Technical articles, learning notes, and reflections will appear here.

Publishing channels

CSDN ↗Zhihu ↗GitHub ↗

The notes below are hosted on this site. External publication status follows the actual CSDN / Zhihu posts.

Engineering notes

01FA3: deterministic SWA backward schedulingShorten the dQ semaphore dependency chain inside the fused kernel.Full backward · 3.98×02FC1: compute–communication contentionJointly tune NCCL CTA and DeepGEMM SM budgets in a dual-stream, four-H200 proxy.Proxy joint latency −12.80%03Deterministic MoE router GEMMThree backends for small-M decode with CUDA Graph preflight and bitwise checks.Router latency −73.39%04R3: MoE training–rollout route consistencyConnect vLLM route capture to Megatron replay, with development validation at 128 GPUs.Route mismatch → 0 05LLMQRT: quantized inference runtimeA PyTorch Extension runtime for two-H200, 32B quantized inference and tensor parallelism.Output throughput +76.8%

Research

AAAI 2027 · Submitted

From Structure to Preference: Token Weighting for Chain-of-Thought Distillation in Large Language Models

Co-author; model training, hyperparameter tuning and vLLM TP=4 evaluation.