仓库/Skill 库

非自回归的 System 1 决策引擎。单次前向传播即可对任意文本进行类型化选择、打分和 yes/no 决策，支持 100+ 语言，并通过路由按请求选择合适的 checkpoint。Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.

FinGPT:开源金融大语言模型!革命性 🔥 已在 HuggingFace 发布训练好的模型。FinGPT: Open-Source Financial Large Language Models! Revolutionize 🔥 We release the trained model on HuggingFace.

中文LLaMA&Alpaca大语言模型+本地CPU/GPU训练部署 (Chinese LLaMA & Alpaca LLMs)

复旦大学开源的工具增强对话语言模型An open-source tool-augmented conversational language model from Fudan University

面向本地部署的高速 LLM 服务High-speed Large Language Model Serving for Local Deployment

提供 AI 应用和模型服务的最简方式 —— 构建模型推理 API、任务队列、LLM 应用、多模型 pipeline 等。The easiest way to serve AI apps and models - Build Model Inference APIs, Job queues, LLM apps, Multi-model pipelines, and more!

开放式前沿智能Open Frontier Intelligence

中文LLaMA-2 & Alpaca-2大模型二期项目 + 64K超长上下文模型 (Chinese LLaMA-2 & Alpaca-2 LLMs with 64K long context models)

基于 Qwen3.5 构建的小型 Jev 风格决策模型家族，可自行训练与部署。tiny Jev-like family of decision models built on top of Qwen3.5 you can train and run on your own

面向全模态模型的高效推理框架。A framework for efficient model inference with omni-modality models

用于高性能 AI 模型服务（vLLM、SGLang）和按需 SSH 访问 GPU 实例的 GPU 集群管理器。A GPU cluster manager for high-performance AI model serving (vLLM, SGLang) and on-demand SSH-accessible GPU instances.

高质量且极速的预训练深度学习模型与 demo。Pre-trained Deep Learning models and demos (high quality and extremely fast)

Fengshenbang-LM(封神榜大模型)是IDEA研究院认知计算与自然语言研究中心主导的大模型开源体系，成为中文AIGC和认知智能的基础设施。

SOTA 低比特 LLM 量化（INT8/FP8/MXFP8/INT4/MXFP4/NVFP4）与稀疏化方案；面向 PyTorch、TensorFlow 与 ONNX Runtime 的领先模型压缩技术SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

一款围绕该模型构建的、面向 Apple silicon 的本地推理引擎。A local inference engine for Apple silicon, built around the model.

基于 2x DGX Spark 与 TensorFold 的 GLM-5.3-Flash EXL3。GLM-5.3-Flash EXL3 on 2x DGX Spark with TensorFold

DeepSeek v4 Flash EXL3 运行于单台 DGX SparkDeepSeek v4 Flash EXL3 on one DGX Spark

GLM-5.3 Flash EXL3，针对 2x DGX Sparks。GLM-5.3 Flash EXL3 for 2x DGX Sparks

用于解析日期和时间的小型神经模型A small neural model for parsing date & time

JevK5：TypeSafe Jev 的开源权重替代方案。单次前向即可输出带概率的类型化决策；权重与代码均采用 Apache-2.0 协议。JevK5: open-weight alternative to TypeSafe Jev. Typed decisions with probabilities in one forward pass; Apache-2.0 weights and code.

一个小型开放决策模型：state + 类型化问题 → 校准概率。在 Qwen3.5 上的 Jev / System One 复现。A small open decision model: state + typed questions -> calibrated probabilities. A Jev / System One re-creation on Qwen3.5.

基于 PyTorch 的韩语 LLM 实现与训练项目PyTorch 기반 한국어 LLM 구현 및 학습 프로젝트

基于 Jev API 的开放决策模型：通过前向传递给出带概率的类型化答案（yes/no、choice、score、multi），无需生成。基于 Qwen3.5-4B + LoRA，单卡 GPU。Open decisions model with Jev's API: typed answers (yes/no, choice, score, multi) with probabilities from forward passes, no generation. Qwen3.5-4B + LoRA, one GPU.

运行 Ollama 本地 LLM 服务的 Docker 镜像。默认安全，所有 API 请求需 Bearer token（首次启动时自动生成）。OpenAI 兼容 API。支持首次启动模型预拉取、NVIDIA GPU (CUDA) 加速和持久化模型存储。多架构：amd64、arm64。Docker image to run an Ollama local LLM server. Secure by default, all API requests require a Bearer token (auto-generated on first start). OpenAI-compatible API. Supports first-start model pre-pull, NVIDIA GPU (CUDA) acceleration, and persistent model storage. Multi-arch: amd64, arm64.

Tox21 多终点毒性预测：可复现的研究与推理制品（冻结模型、FastAPI 服务、已审计的安全机制）Tox21 multi-endpoint toxicity prediction: a reproducible research and inference artifact (frozen model, FastAPI service, audited security)

面向生产环境的、有内存约束的跨模型 KV-cache 传输，具备受保护的回退机制与可复现研究工具链。Production-oriented, memory-bounded cross-model KV-cache transfer with guarded fallback and reproducible research tooling.

混合神经符号 AI：Llama 3.2 1B + 精确数学 + 经验证事实，807 MB 即时 CPU 原生回答，无需 GPU。封装 .aef 分发。Hybrid neuro-symbolic AI: Llama 3.2 1B + exact math + verified facts — instant CPU-native answers in 807 MB, no GPU. Sealed .aef distribution

论文《Controllable molecular graph generation from natural-language chemical constraints》（基于自然语言化学约束的可控分子图生成）的代码仓库。This is repository for "Controllable molecular graph generation from natural-language chemical constraints"

Experimental Qwen3.5-derived 752M LLM：基于 Qwen3.5 的实验性 752M 参数 LLM，采用 CPT + SFT 两阶段流程，面向编码、技术推理与指令跟随。Experimental Qwen3.5-derived 752M LLM: a two-phase CPT + SFT pipeline for coding, technical reasoning, and instruction following.

MAI-Code 模型的官方仓库，用于发布 MAI-Code 模型版本与更新，并通过 issue 与反馈与开发者社区互动。Official repo for MAI-Code models, where we publish MAI-Code model releases and updates and engage with developer community on issues and feedback.

🌐 通过 Geo-Llama 利用几何深度学习增强语言理解，结合 conformal manifolds 和递归等距变换提升 AI 模型性能。🌐 Enhance language understanding through geometric deep learning with Geo-Llama, leveraging conformal manifolds and recursive isometries for improved AI models.

RIFT — Race-state Inference From Telemetry。一项可复现研究实现，仅基于位置遥测重建具备事件感知的关联性比赛状态，涵盖路线进度、共享事件标识、检查点通过、物理顺序、前车关系、间距与间隔。RIFT — Race-state Inference From Telemetry. A reproducible research implementation for reconstructing occurrence-aware relational race state from positional telemetry alone, including route progress, shared occurrence identity, checkpoint crossings, physical order, car-ahead relations, gaps and intervals.

目标：将 State Space Models（SSM）/ Mamba 用于医学时间序列分析；总体研究问题：Mamba 模型能否提升基于步态信号检测帕金森病的分类模型性能: To use State Space Models (SSM)/ Mamba for medical time series analysis General research question: Can Mamba models improve the performance of classification models trained to detect Parkinson's disease using gait signals.