Stars
Accelerating MoE with IO and Tile-aware Optimizations
Allow torch tensor memory to be released and resumed later
Ring attention implementation with flash attention
微舆：人人可用的多Agent舆情分析助手，打破信息茧房，还原舆情原貌，预测未来走向，辅助决策！从0实现，不依赖任何框架。
🔥 LLM-powered GPU kernel synthesis: Train models to convert PyTorch ops into optimized Triton kernels via SFT+RL. Multi-turn compilation feedback, cross-platform NVIDIA/AMD, Kernelbook + KernelBench
slime is an LLM post-training framework for RL Scaling.
gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI
SGLang is a fast serving framework for large language models and vision language models.
verl: Volcano Engine Reinforcement Learning for LLMs
My learning notes for ML SYS.
This email knows how long you have been reading it.
Parallel Programming Design Assignment for Nankai University, the topic chosen is the default Gaussian Elimination. 南开并行程序设计作业，默认高斯消去选题。
A tool which is uses to remove Windows Defender in Windows 8.x, Windows 10 (every version) and Windows 11.
A hex editor for WeChat/QQ/TIM - PC版微信/QQ/TIM防撤回补丁（我已经看到了，撤回也没用了）
Open source, compact, and material designed cursor set.
LiteLoaderQQNT 插件 - 轻量工具箱 —— 轻量 · 优雅 · 高效
MMSA is a unified framework for Multimodal Sentiment Analysis.
多模态情感分析——基于BERT+ResNet的多种融合方法
Shortcuts for Siri using ChatGPT API gpt-3.5-turbo & gpt-4 model, supports continuous conversations, configure the API key & save chat records. 由 ChatGPT API gpt-3.5-turbo & gpt-4 模型驱动的智能 Siri，支持连续…
SSR 去广告ACL规则/SS完整GFWList规则/Clash规则碎片，Telegram频道订阅地址