Skip to content
View zhengwenze's full-sized avatar
🌴
On vacation
🌴
On vacation

Block or report zhengwenze

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
zhengwenze/README.md

Hi ~ I'm Wenze Zheng

AI Infra 工程师 · 专注大模型推理引擎与推理服务评测

zhengwenze visitor counter

Python PyTorch CUDA Triton vLLM FlashAttention FastAPI Docker Linux

ABOUT ME

  • 🏙️ 北京,AI Infra 工程师,方向为大模型推理引擎与推理服务评测
  • � 自研轻量 vLLM 风格推理引擎,研究 GPU 在线调度(Prefill/Decode 策略、Paged KV Cache、FlashAttention、SSE 流式)
  • 📊 自研推理服务评测平台,测量 TTFT / TPOT / 吞吐 / Goodput,产出可复核的实验数据
  • 🧩 喜欢把复杂的推理问题拆成可复现、可度量的工程实验

FEATURED PROJECTS

轻量 vLLM 风格推理引擎,二次开发在线请求调度:Prefill First / Decode First / Bounded Decode First 三策略、逐 Token SSE 流式、请求取消与背压;在 RTX 4060 / Qwen3-0.6B 上实测调度策略对 TTFT / TPOT 的影响。

Python PyTorch CUDA Triton FlashAttention SSE

面向 OpenAI 兼容推理服务的评测平台:流式压测、实验有效性校验(SLO / Goodput / GPU 隔离),采集 vLLM 与 NVML 遥测,产出可复核的原始数据、报告与图表。

Python vLLM Prometheus NVML pytest

CONTACT

Email GitHub

Pinned Loading

  1. zwz-infer-scope zwz-infer-scope Public

    zwz-infer-scope 在现有性能 Benchmark 基础上补齐真正的模型评测

    Python

  2. zwz-vllm-lab zwz-vllm-lab Public

    Forked from GeeeekExplorer/nano-vllm

    zwz-vllm-lab

    Python