Skip to content
View CUHKSZzxy's full-sized avatar
:octocat:
Focusing
:octocat:
Focusing
42 followers · 27 following

Achievements

Achievement: Pair Extraordinairex2Achievement: QuickdrawAchievement: Pull Sharkx2Achievement: Arctic Code Vault Contributor

Achievements

Achievement: Pair Extraordinairex2Achievement: QuickdrawAchievement: Pull Sharkx2Achievement: Arctic Code Vault Contributor

Block or report CUHKSZzxy

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
CUHKSZzxy/README.md

Hi there 👋

Stats

Open-source Work

Focused on LLM/VLM serving systems.

Selected contributions:

#PR Status Summary
LMDeploy GLM-5.2 optimizations 🟣 #4896 bounded DSA prefill score memory, short DSA prefill fast path
#4853 CUDA all-reduce, BF16 router GEMM, PDL
#4827 DSA index scoring, Q/KV-A projection fusion, LM-head sharding, etc.
LMDeploy model support 🟢/🟣 #4780 Intern-S2-Preview TS (397B); #4737 GLM-5.2 (753B)
#4575 Intern-S2-Preview (35B-A3B); #4411 Qwen3-Omni (30B-A3B)
#4318 Intern-S1-Pro (1T-A22B); #4093 Qwen3-VL (2B - 235B-A22B)
#3863 GLM-4.5 (106B-A12B - 355B-A32B); #3846 GLM-4.1V (9B)
#3315 Qwen3/MoE (0.6B - 235B-A22B); #3194 Qwen2.5-VL (3B - 72B)
#3149 DeepSeek-VL2 (3B - 27B).
LMDeploy #4582 🟣 Add OpenAI Responses-compatible API endpoint.
LMDeploy #4563 🟣 FP8 KV-cache quantization (quant policy, cache engine, attention kernels).
LMDeploy #4531 🟣 Mixed-modality serving (multimodal inputs, VLM preprocessing, chat templates).
LMDeploy #4360 🟣 Video input serving (video loader, VL media, model inputs).
LMDeploy #3534 🟣 Serving metrics (metrics processor, Prometheus, Grafana).
vLLM model support 🟣 #42705 Intern-S2-Preview (35B-A3B) (co-authored); #33636 Intern-S1-Pro (1T-A22B).
SGLang model support 🟣 #9299 Intern-S1-mini (9B); #8350 Intern-S1 (241B) (co-authored).

Status: 🟣 merged, 🟢 open.

Complete contributions:

Pinned Loading

  1. InternLM/lmdeploy InternLM/lmdeploy Public

    LMDeploy is a toolkit for compressing, deploying, and serving LLMs.

    Python 8k 740

  2. vllm-project/vllm vllm-project/vllm Public

    A high-throughput and memory-efficient inference and serving engine for LLMs

    Python 91.3k 21.9k