Projects

### Overview: EvalScope is a comprehensive and customizable framework designed for efficient evaluation and benchmarking of large language models (LLM), vision language models (VLM), and more. It supports multiple evaluation benchmarks and integrates various evaluation backends for seamless performance testing. ### Key Features: - **📚 Comprehensive Evaluation Benchmarks**: Built-in benchmarks such as MMLU, C-Eval, GSM8K, and more. - **🧩 Multi-modal and Multi-domain Support**: Evaluates models across various domains and types including LLM, VLM, embedding, reranker, and more. - **🧭 System One Evaluation**: Evaluates text-choice decision models with structured probabilities and confidence. - **🚀 Multi-backend Integration**: Seamlessly integrates OpenCompass, VLMEvalKit, and RAGEval for diverse evaluation needs. - **🤖 Agent Evaluation Mode**: Drives benchmarks inside a controlled multi-turn AgentLoop with pluggable strategies and Docker sandbox. - **⚡ Inference Performance Testing**: Provides powerful model service stress testing tools for metrics like TTFT and TPOT. - **📊 Interactive Reports**: Offers a Web Dashboard for model comparison, overview, and detailed analysis. - **⚔️ Arena Mode**: Supports multi-model battles for intuitive ranking and evaluation. - **🔧 Highly Extensible**: Allows developers to add custom datasets, models, and evaluation metrics. ### Why It Matters: EvalScope simplifies the evaluation and benchmarking process, making it easier and more efficient to compare and analyze the performance of large language and vision models across various domains and types.

Overview: This repository contains all the lecture notes, slides, and assignments from the CS229: Machine Learning course taught by Stanford University. It serves as a comprehensive resource for anyone interested in machine learning. Key features: - Lecture notes and slides - Assignments and quizzes - Course materials for both the CS229: Machine Learning course and the CS229: Machine Learning (Summer 2019) edition Tech stack: Jupyter Notebook Why it matters: This repository is an invaluable resource for students and professionals looking to deepen their understanding of machine learning. It provides a structured way to learn from the course materials and stay updated with the latest developments in the field.

### Overview: This repository is a curated list of Generative AI projects, tools, artworks, and models. It serves as a comprehensive resource for developers and enthusiasts interested in the latest advancements in Generative AI. ### Key Features: - Generative AI history, timelines, maps, and definitions - Ethical, philosophical questions, and discussions about Generative AI - Critical views about Generative AI - Generative AI processes and artifacts - Generative AI tools directories - Courses and educational materials - Human-AI interaction - Papers collection - Online tools and applications ### Tech Stack: GitHub stars: 3555 ### Why it Matters: This repository provides a wealth of information on Generative AI, making it an essential resource for anyone looking to stay updated with the latest developments in the field.

Overview: MNE-Python is an open-source Python package designed for exploring, visualizing, and analyzing neurophysiological data such as MEG, EEG, sEEG, and ECoG. Key features: - Data input/output - Preprocessing - Visualization - Source estimation - Time-frequency analysis - Connectivity analysis - Machine learning - Statistics Tech stack: Python Why it matters: MNE-Python simplifies the process of analyzing neurophysiological data, making it accessible to researchers and developers alike.

Overview: Hummingbird is a library that compiles trained ML models into tensor computations, enabling faster inference with the benefits of neural network optimizations, native hardware acceleration, and a unified platform for traditional and neural network models. Key features: - Converts traditional ML models into PyTorch, TorchScript, ONNX, and TVM models. - Supports a variety of ML models and featurizers including scikit-learn, LightGBM, and XGBoost. - Provides a Sklearn API-compatible inference API. - Supports multiple tree conversion strategies, including GEMM. Tech stack: Python Why it matters: Hummingbird simplifies the process of integrating traditional ML models with neural network frameworks, leveraging the strengths of both worlds for faster and more efficient inference.

Overview: This repository contains essential components for text processing using PyTorch, including datasets, data loaders, transformations, models, and vocabularies. Key features: - `torchtext.datasets`: Raw text iterators for common NLP datasets - `torchtext.data`: Basic NLP building blocks - `torchtext.transforms`: Basic text-processing transformations - `torchtext.models`: Pre-trained models - `torchtext.vocab`: Vocab and Vector related classes and factory functions - `examples`: Example NLP workflows with PyTorch and torchtext library Tech stack: Python Why it matters: This repository is crucial for developers working with natural language processing (NLP) tasks, providing a comprehensive toolkit for text processing using PyTorch.

### Overview: SiamMask is an advanced framework designed for real-time online object tracking and segmentation. This project introduces a unified approach that significantly improves efficiency and accuracy in tracking and segmentation tasks. ### Key features: - **Real-time Tracking**: SiamMask achieves real-time performance, making it suitable for applications requiring continuous tracking. - **High Accuracy**: The framework provides high accuracy in object tracking and segmentation, ensuring reliable results. - **Versatile Models**: It supports both training and inference, allowing for flexibility in deployment and experimentation. ### Tech stack: - **Python**: The primary programming language used for implementing the framework. - **PyTorch**: Used for implementing the neural network models within the framework. ### Why it matters: SiamMask is significant because it offers a comprehensive solution for online object tracking and segmentation tasks, making it a valuable tool for researchers and practitioners in computer vision. The unified approach and real-time capabilities make it particularly useful in applications where speed and accuracy are critical. ### Repository: - **Repository**: `foolwood/SiamMask` - **GitHub Stars**: 3551 - **Primary Language**: Python - **GitHub Description**: "CVPR19/TPAMI23] SiamMask: A Framework for Fast Online Object Tracking and Segmentation" ### Contents: - **Environment Setup**: Ensures the correct environment is set up for running the code. - **Demo**: Provides a step-by-step guide to setting up and running the demo. - **Testing Models**: Includes instructions on how to test the models with specific configurations. ### Environment

Overview: ML Workspace is an all-in-one web-based IDE designed specifically for machine learning and data science, providing a seamless environment for rapid development and deployment. Key features: - Jupyter Notebook - Visual Studio Code web-based IDE - Pre-installed with popular data science libraries and tools - Full Linux desktop GUI accessible via web browser - Seamless Git integration - Tensorboard and Netdata integration - Access from anywhere via Web, SSH, or VNC - Deployable on Mac, Linux, and Windows via Docker Tech stack: Jupyter Notebook, Visual Studio Code, Docker Why it matters: ML Workspace simplifies the development process by providing a comprehensive and optimized environment for machine learning and data science projects, making it ideal for developers looking to quickly and efficiently build and deploy ML solutions on their own machines.

### Overview: pomegranate is a fast, flexible, and easy-to-use probabilistic modeling library in Python, designed for developers and researchers alike. It provides a modular implementation that allows for the seamless integration of various distributions and models. ### Key features: - **Modular Implementation**: Supports the addition of any distribution or model as a probability distribution. - **Flexible Bayesian Networks**: Can be easily integrated into mixture models. - **Heterogeneous Mixture Models**: Combines multiple distributions to create complex models. - **Hidden Markov Models**: Can be used as classifiers over sequences. - **GPU Support**: Utilizes PyTorch's GPU capabilities for faster computations. ### Tech stack: PyTorch ### Why it matters: pomegranate is a powerful tool for probabilistic modeling, offering flexibility and speed that are not found in other libraries. Its modular design allows for easy integration of new features and models, making it a valuable addition to any developer's toolkit.

### Overview: 本书是《神经网络与深度学习：案例与实践》第二版，详细介绍了使用 PyTorch 实现关键算法、模型和工程实践。内容涵盖从基础线性模型到复杂模型如注意力机制和大语言模型的全面介绍，适合希望动手学习深度学习的学生和工程师。 ### Key features: - **PyTorch 实践**：从零开始实现关键算法和模型。 - **逐章实现说明**：每章提供详细的实现说明和测试。 - **配套代码**：提供从基础算子到复杂模型的完整工程化版本，便于复用。 - **数据准备**：提供数据集下载和准备说明。 - **章节目录**：详细列出每章主题和主要实践内容。 ### Tech stack: 使用 Jupyter Notebook 实现。 ### Why it matters: 本书结合了理论讲解和实践操作，帮助读者从零开始构建和优化深度学习模型，适用于希望系统学习深度学习实践的学生和工程师。

Overview: Farfalle is an open-source AI-powered search engine designed for self-hosting with either local or cloud language models. It supports a variety of search providers and custom LLMs through LiteLLM. Key features: - Supports local language models (llama3, gemma, mistral, phi3) and custom LLMs through LiteLLM. - Answers questions using cloud models (OpenAI/gpt4-o, OpenAI/gpt3.5-turbo, Groq/Llama3). - Answers questions with local models (llama3, mistral, gemma, phi3). - Answers questions with any custom LLMs through LiteLLM. - Provides chat history and expert search capabilities. - Supports chat with local files. Tech stack: TypeScript Why it matters: Farfalle offers a flexible and customizable solution for developers looking to integrate AI-powered search capabilities into their projects. It allows for easy integration of various language models and search providers, making it a versatile tool for developers and content creators alike. Repository: rashadphz/farfalle Primary language: TypeScript GitHub stars: 3542 GitHub description: 🔍 AI search engine - self-host with local or cloud LLMs README excerpt: # Farfalle Open-source AI-powered search engine. (Perplexity Clone) Run local LLMs (llama3, gemma, mistral, phi3), custom LLMs through LiteLLM, or use cloud models (Groq/Llama3, OpenAI/gpt4

Overview: WorkBuddy Guide is a practical, open-source guide designed to help developers master WorkBuddy through real-world workflows. It offers tutorials, real-world cases, skills, MCP, automation, and multi-agent practices. Key features: - Tutorials on installing and using WorkBuddy - Real-world case studies covering various tasks and industries - Skills and MCP (Mastery Certification Program) for advanced users - Case studies on team development and deployment - Case submissions and community contributions for further development Tech stack: TypeScript Why it matters: This guide provides a comprehensive resource for developers looking to enhance their skills and knowledge in using WorkBuddy effectively. Repository: AlephAITech/WorkBuddyGuide GitHub stars: 3541 GitHub description: A practical, open-source guide to mastering WorkBuddy through real-world workflows. 开源的 WorkBuddy 实战蓝皮书：教程、真实工作流、Skills、MCP、自动化与多智能体实践。

Overview: Scikit-LLM seamlessly integrates powerful language models like ChatGPT into scikit-learn, enhancing text analysis tasks. Key features: - Zero-shot text classification using GPT - Integration with scikit-learn for data preprocessing and model training - Support for various LLM models, including GPT-4 Tech stack: Python Why it matters: Scikit-LLM simplifies the integration of large language models into machine learning workflows, making advanced text analysis tasks more accessible and efficient.

### Overview: Mesh-LLM is a distributed AI/LLM platform that enables the pooling of GPUs and memory across machines. It exposes an OpenAI-compatible API at `http://localhost:9337/v1`, allowing users to share compute resources privately or publicly to power their agents and chat. The system supports peer discovery, routing, and plugin integration, making it highly scalable and flexible. ### Key Features: - **Mesh-LLM Host and Console**: Provides a unified interface for managing and serving models. - **Skippy**: Manages model management, inference, and split execution. - **Public Mesh**: Enables sharing compute resources on a public network. - **Private Mesh**: Allows for private sharing of compute resources. - **Model Routing**: Routes models to the most optimal location based on network conditions. - **Model Management**: Supports native inference and model management. - **Serving**: Provides an OpenAI-compatible API for model serving. - **Web Console**: Offers a user-friendly interface for monitoring and managing the system. ### Tech Stack: - **Rust**: The primary language used for the backend and frontend development. - **Cargo**: The build system for Rust projects. - **Skippy**: The core component responsible for model management, inference, and routing. ### Why It Matters: Mesh-LLM is crucial for anyone looking to leverage distributed AI/LLM capabilities without the need for complex infrastructure setup. It simplifies the process of sharing and managing compute resources, making it accessible to developers and researchers alike. The ability to easily join a public mesh or create a

Overview: Refact is an AI agent designed to handle engineering tasks end-to-end, integrating with developers' tools, planning, executing, and iterating until a successful result is achieved. Key features: - Integrates with IDE chat - Autonomous agent workflows - Tool-powered development Tech stack: Rust Why it matters: Refact is a powerful tool for developers looking to automate and streamline their workflow, making it easier to handle complex engineering tasks efficiently. Repository: smallcloudai/refact Primary language: Rust GitHub stars: 3533

### Overview: TensorFlow Hub is a library that allows developers to reuse parts of TensorFlow models for transfer learning. It provides a simple interface to load and use pre-trained models, making it easier to build and deploy machine learning applications. ### Key Features: - **Model Reusability**: Reuse pre-trained models by loading them via `hub.load()`. - **Model Loading**: Easily load models from TensorFlow Hub using `hub.load`. - **Model Versioning**: Support for multiple versions of the same model. ### Tech Stack: Python ### Why It Matters: TensorFlow Hub simplifies the process of reusing pre-trained models, making it easier to build and deploy machine learning applications. It reduces the need for retraining models from scratch, saving time and resources.

### Overview: The Image Restoration Toolbox (KAIR) is a comprehensive project that provides training and testing codes for various image restoration techniques such as DPIR, USRNet, DnCNN, FFDNet, SRMD, DPSR, MSRResNet, ESRGAN, BSRGAN, SwinIR, VRT, and RVRT. The toolbox is designed to help developers and researchers in the field of computer vision to improve their image restoration capabilities. ### Key Features: - **DPIR (Deep Plug-and-Play Image Restoration)**: A state-of-the-art image restoration method that leverages deep learning to enhance images. - **USRNet (Unsupervised Super-Resolution Network)**: A network that can upscale images without paired training data, making it highly effective for real-world applications. - **DnCNN (Deep Neural Network for Denoising)**: A convolutional neural network specifically designed for image denoising tasks. - **FFDNet (Fast Feature-Domain Network)**: A network that operates in the feature domain to enhance image quality. - **SRMD (Super-Resolution with Multiple Dictionaries)**: A method that uses multiple dictionaries to improve the resolution of images. - **DPSR (Deep Plug-and-Play Super-Resolution)**: A super-resolution technique that can be used for both image and video restoration. - **MSRResNet (Multi-Scale Residual Network)**: A network that enhances image quality by learning multiple scales. - **ESRGAN (Enhanced Super-Resolution Generative Adversarial Network

### Overview: This repository contains the source code for my personal column "Machine Learning Without Tears" on Zhihu. It includes projects in Natural Language Processing (NLP) and Computer Vision, such as text generation, machine translation, deep convolutional GANs, and other practical code examples. ### Key features: - Text generation using LSTM models - Word2Vec model for skip-gram algorithm - Lyrics generation using RNN - Seq2Seq model for sorting words - Image denoising using autoencoder - MNIST handwritten digit generation using GANs - CIFAR image classification using CNNs - MNIST digit generation using DCNNs - Batch normalization implementation - Machine translation model using Seq2Seq - Attention mechanism and BiRNN in machine translation - Sentiment analysis model using DNN, LSTM, and CNN - Image style transfer model using TensorFlow ### Tech stack: Python 3.6 ### Why it matters: This repository provides a comprehensive collection of practical projects and code examples in Natural Language Processing and Computer Vision, making it useful for developers looking to enhance their skills in these areas.

### Overview: PyTracking is a comprehensive Python framework for visual object tracking and video object segmentation, built on PyTorch. ### Key features: - **TaMOs (WACV 2024)**: A generico object tracker for multiple generic objects. - **RTS (ECCV 2022)**: Robust end-to-end trainable segmentation-centric pipeline. - **ToMP (CVPR 2022)**: Segmentation-centric pipeline with instance localization. - **KeepTrack (ICCV 2021)**: Efficient visual tracking with exemplar transformers. - **LWL (ECCV 2020)**: Lightweight, efficient tracker. - **KYS (ECCV 2020)**: Key-Value instance tracker. - **PrDiMP (CVPR 2020)**: Pruning-based instance tracker. - **DiMP (ICCV 2019)**: Deep instance-based tracker. - **ATOM (CVPR 2019)**: Advanced tracking model. ### Tech stack: PyTorch. ### Why it matters: PyTracking provides a robust and efficient toolkit for visual object tracking and segmentation, making it a valuable resource for researchers and developers in the field. Repository: visionml/pytracking Primary language: Python GitHub stars: 3520 GitHub description: Visual tracking library based on PyTorch.

Overview: ManimML is a project focused on creating animations and visualizations of common machine learning concepts using the Manim Community Library. Key features: - Animations of neural networks - Visualizations of convolutional neural networks - Demonstration of max pooling - Explanation of activation functions - Introduction to neural network dropout Tech stack: Python Why it matters: ManimML aims to provide a comprehensive set of visualizations for machine learning concepts, making it easier for users to understand complex ideas in a more intuitive way.