Skip to content
View zhuzilin's full-sized avatar
🛏️
躺平躺平......
🛏️
躺平躺平......

Block or report zhuzilin

Report abuse

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

Report abuse
zhuzilin/README.md

AI is the next chip.

Hey, I'm zhuzilin, an engineer driven by curiosity.

My main focus is on MLSys.

  • You can ask me about deep learning frameworks. I am contributor to many tools like pytorch, tensorflow and horovod.
  • I am a LLM believer and was really lucky to get hands dirty on training them @WeChat, from pretraining from scratch to sft and rlhf, along with writing training frameworks for those.
  • Currently working on the training framework for RL + LLM at @zhipuAI.

I'm also interested in JavaScript engine. I've read the es5 spec to write es and helped fixed bugs in the early stage of oven-sh/bun.

Avatar is Shoyo Hinata, from Haikyu!!.


我是 zhuzilin,一个由兴趣驱动的工程师~

我的主要精力放在 MLSys 领域。

  • 我比较了解深度学习训练框架,是 pytorch, tensorflow, horovod 等工具的 contributor。
  • LLM 信徒,之前在微信大模型团队打工的过程中,有幸深入接触过 LLM 训练的各个环节,不管是从零预训练,还是 sft 与 rlhf,以及写用来做这些事的训练框架。
  • 目前在智谱做 RL 训练框架。

我对 JavaScript 引擎也比较感兴趣。读过 spec,写过解释器(es),还给早期的 oven-sh/bun 提过一些 bugfix。

头像是日向翔阳,《排球少年》。

Pinned Loading

  1. THUDM/slime THUDM/slime Public

    slime is an LLM post-training framework for RL Scaling.

    Python 7.3k 1k

  2. ring-flash-attention ring-flash-attention Public

    Ring attention implementation with flash attention

    Python 1k 99

  3. OpenRLHF/OpenRLHF OpenRLHF/OpenRLHF Public

    An Easy-to-use, Scalable and High-performance Agentic RL Framework based on Ray (PPO & DAPO & REINFORCE++ & VLM & TIS & vLLM & Ray & Async RL)

    Python 9.7k 976

  4. pytorch/pytorch pytorch/pytorch Public

    Tensors and Dynamic neural networks in Python with strong GPU acceleration

    Python 101k 28.2k

  5. es es Public archive

    A JavaScript interpreter from scratch, supporting ES5 syntax.

    C++ 30 6

  6. oven-sh/bun oven-sh/bun Public

    Incredibly fast JavaScript runtime, bundler, test runner, and package manager – all in one

    Rust 93.6k 4.7k