Projects

### Overview: FSRS4Anki is a modern spaced-repetition scheduler designed specifically for Anki, based on the Free Spaced Repetition Scheduler algorithm. It optimizes the scheduling of flashcards to enhance retention and learning efficiency. ### Key Features: - **Scheduler**: Replaces Anki's built-in scheduler to schedule cards according to the FSRS algorithm. - **Optimizer**: Uses machine learning to learn your memory patterns and find parameters that best fit your review history. - **Add-on Compatibility**: FSRS can be used alongside Anki add-ons like Review Heatmap, Advanced Browser, and The KING of Button Add-ons, provided they do not affect the scheduling process. ### Tech Stack: Jupyter Notebook ### Why It Matters: FSRS4Anki leverages the Free Spaced Repetition Scheduler algorithm to enhance the scheduling of flashcards, making the learning process more efficient and effective. The optimizer learns from your review history to find the best parameters for your learning style, ensuring that you retain information more effectively. ### Repository: - **Repository**: open-spaced-repetition/fsrs4anki - **Stars**: 4110 - **Description**: A modern Anki custom scheduling based on the Free Spaced Repetition Scheduler algorithm - **License**: MIT License - **Contributors**: See the README for contributors ### Stargazers Over Time: - **Stargazers**: 4110 ### Acknowledgements: - The FSRS algorithm was inspired by Maimemo, a research project that focuses on data-driven

### Overview: OpenClaw Mission Control is an AI Agent Orchestration Dashboard designed for managing and coordinating multi-agent collaboration via OpenClaw Gateway. It was an experimental platform that has been superseded by Wardn AI, which offers a cleaner, more scalable architecture, active maintenance, and better long-term support. ### Key Features: - AI Agent Orchestration Dashboard - Multi-Agent Collaboration - OpenClaw Gateway Integration ### Tech Stack: TypeScript ### Why it Matters: OpenClaw Mission Control was an experimental operations platform for OpenClaw. As the project evolved, its architecture and long-term direction diverged significantly from what Mission Control was originally designed for. Wardn AI provides a cleaner, more scalable architecture, active maintenance, and better long-term support for AI agent orchestration and operations. ### Repository: abhi1693/openclaw-mission-control ### Primary Language: TypeScript ### GitHub Stars: 4102 ### GitHub Description: AI Agent Orchestration Dashboard - Manage AI agents, assign tasks, and coordinate multi-agent collaboration via OpenClaw Gateway. ### GitHub Description (archived): This repository has been archived and is no longer maintained. Development has moved to Wardn AI, which supersedes OpenClaw Mission Control with a significantly improved architecture, better maintainability, and continued active development. If you are looking to build, deploy, or operate AI agents, you should use Wardn AI instead of this repository. Please migrate to Wardn AI for all future deployments and development. Repository: https://github.com/ab

Overview: SSD is a unified framework for object detection that has been introduced in this research paper. This repository contains a TensorFlow re-implementation of the original Caffe code, focusing on VGG-based SSD networks with 300 and 512 inputs. The project is modular, making it easy to implement and train other SSD variants. Key features: - A unified framework for object detection - Modular architecture for easy implementation and training of other SSD variants - Supports Pascal VOC datasets (2007 and 2012) - TensorFlow re-implementation of original Caffe code Tech stack: TensorFlow Why it matters: This project is crucial for developers looking to implement object detection using TensorFlow, as it provides a comprehensive and modular solution for SSD networks. The TensorFlow implementation ensures compatibility with existing Caffe models, making it easier to integrate with existing projects and datasets. Repository: balancap/SSD-Tensorflow Primary language: Jupyter Notebook GitHub stars: 4101 GitHub description: Single Shot MultiBox Detector in TensorFlow

Overview: Cairn is a general-purpose AI state-space search engine designed for solving problems with a clear starting point, a clear success condition, and an unknown path in between. It has been validated on autonomous penetration testing and is built on a Blackboard Architecture with an explicit fact-intent graph. Key features: Cairn is built for AI penetration testing, vulnerability research, and other problems with a defined origin and goal but an unknown path in between. It uses a Blackboard Architecture with three primitives: Facts, Intent, and Hint. Agents run an OODA loop, and tasks are generated dynamically from the graph's current state. Tech stack: Cairn is written in Python and uses a Blackboard Architecture with an explicit fact-intent graph. Why it matters: Cairn is designed to solve problems with a clear starting point and an unknown path in between, making it a powerful tool for tasks like AI penetration testing, vulnerability research, and other similar problems.

### Overview SuperGlue is a research project that combines a Graph Neural Network with an Optimal Matching layer to perform feature matching on sparse image features. This project was presented at CVPR 2020 and is now available for use with the provided pre-trained weights. ### Key Features - **SuperPoint**: A lightweight feature detector that extracts keypoints and descriptors. - **SuperGlue**: A Graph Neural Network combined with an Optimal Matching layer for feature matching. - **Pretrained Weights**: Available for indoor (ScanNet) and outdoor (MegaDepth) models. ### Tech Stack - Python 3.5+ - PyTorch 1.1+ - OpenCV 3.4 (4.1.2.30 recommended for best GUI interaction) - Matplotlib 3.1+ ### Why It Matters SuperGlue is a powerful tool for performing feature matching on image pairs, making it useful in various applications such as visual localization, structure from motion (SfM), and more. The pre-trained weights provided by the project can be used to quickly run feature matching on images without the need for extensive training. ### Repository The repository is hosted on GitHub at [magicleap/SuperGluePretrainedNetwork](https://github.com/magicleap/SuperGluePretrainedNetwork). It includes PyTorch code and pretrained weights for running the SuperGlue matching network. ### Dependencies - Python 3.5+ - PyTorch 1.1+ - OpenCV 3.4 (4

Overview: Answer me with HTML is an agent skill that converts complex HTML questions into a readable one-page HTML response, replacing the traditional wall of text with a concise and easily digestible format. Key features: - Converts complex HTML questions into a one-page HTML response - Uses a model that writes only Markdown drafts, resulting in significantly fewer tokens and faster response times - Supports installation and integration with various platforms Tech stack: JavaScript Why it matters: This skill provides a more user-friendly and efficient way to receive answers, reducing the amount of tokens the model needs to write and improving the overall response time.

Overview: APM is an open-source, community-driven dependency manager for AI agents, providing a single manifest file to describe and configure all dependencies, ensuring portable and reproducible setups across agents and clients. Key features: - **One manifest file** for every agent's context. - **One command** to reproduce the setup across all clients. - **One policy** to govern what an organization allows. Tech stack: Python Why it matters: APM simplifies the setup and management of AI agent configurations, making it easier for developers to share and reuse dependencies across projects, reducing manual configuration efforts and promoting a more standardized approach to AI development.

### Overview: Baichuan 2 是百川智能推出的新一代开源大语言模型，采用高质量语料训练，性能在多个领域中取得最佳效果。所有版本对学术研究完全开放，开发者可通过邮件申请并获得商用许可。 ### Key features: - **7B Base Model**: 基座模型，适用于通用任务。 - **13B Chat Model**: 适用于对话和问答任务。 - **4bits Quantization**: 4bits量化版本，适合低资源环境。 ### Tech stack: Python ### Why it matters: Baichuan 2 是一个强大的语言模型，适用于各种应用场景，包括但不限于客服、教育、医疗、法律等领域。其性能在多个权威数据集上取得了最佳效果，为学术研究和实际应用提供了有力支持。 Repository: baichuan-inc/Baichuan2 GitHub stars: 4080 GitHub description: A series of large language models developed by Baichuan Intelligent Technology

Overview: This is a 12-chapter, 7,500+ line interactive textbook that teaches you how to build, train, and run a modern language model from absolute scratch. The same architecture behind ChatGPT, Claude, LLaMA, and Mistral. Key features: - Tokenization - Embeddings - Attention mechanism - Training loop - Inference engine Tech stack: Jupyter Notebook Why it matters: After finishing, you won't just know that attention "works". You'll understand the variance argument behind `1/√d_k`. How RoPE captures relative position through rotation. Why pre-norm beats post-norm for deep networks. And exactly where every gradient flows during backpropagation. Repository: raiyanyahya/how-to-train-your-gpt Primary language: Jupyter Notebook GitHub stars: 4077 GitHub description: Build a modern LLM from scratch. Every line commented. Explained like we are five.

Overview: This repository is a curated list of resources on responsible machine learning, including AI governance, assessment, implementation practices, policy, institutional guidance, incidents, accountability, and related research materials. Key features: - AI Governance and Policy Resources - AI Assessment and Implementation Resources Tech stack: Python, Markdown, Git Why it matters: This repository provides essential resources for developers and researchers interested in responsible machine learning, ensuring that AI is developed and implemented ethically and transparently. Repository: jphall663/awesome-machine-learning-interpretability GitHub stars: 4074

### Overview: The Learning to Learn project demonstrates the ability of neural networks to adapt and improve their own learning process. This innovative approach leverages TensorFlow to train and optimize the learning rate of neural networks, making them more efficient and effective. ### Key Features: - **TensorFlow Integration**: Utilizes TensorFlow for training and evaluation. - **Adaptive Learning Rate**: Automatically adjusts the learning rate based on performance metrics. - **Problem-Specific Optimizers**: Supports multiple problems including classification tasks and neural network architectures. ### Tech Stack: Python ### Why It Matters: This project showcases the potential of neural networks to self-optimize, potentially leading to more efficient and effective learning algorithms. It demonstrates the power of combining machine learning with TensorFlow, making it a valuable resource for researchers and developers looking to enhance their models' performance.

### Overview: This repository is a curated collection of resources for quantitative and algorithmic trading, with a strong emphasis on machine learning. It aims to provide a comprehensive guide for those interested in mastering the techniques and tools used in modern financial markets. ### Key Features: - **Books**: A diverse selection of books on financial machine learning, including advanced algorithms and practical applications. - **Online Courses**: A mix of courses from reputable institutions like Udacity and Coursera, focusing on machine learning and reinforcement learning in finance. - **YouTube Videos**: A selection of insightful video tutorials on stock market prediction using deep learning. ### Tech Stack: - Python - TensorFlow - PyTorch - Scikit-learn - Jupyter Notebooks ### Why it Matters: This repository is a valuable resource for anyone looking to enhance their knowledge and skills in quantitative trading. It covers a wide range of topics, from theoretical foundations to practical applications, ensuring a well-rounded understanding of the subject matter.

Overview: FEDML is a unified and scalable machine learning library designed for large-scale distributed training, model serving, and federated learning. It offers a cross-cloud scheduler, TensorOpera AI, enabling the execution of AI jobs on any GPU cloud or on-premise cluster. Key features: - Federated Learning - Distributed Training - Model Serving - Cross-Cloud Scheduling Tech stack: Python Why it matters: FEDML simplifies the deployment and execution of AI models across decentralized and distributed environments, making it easier and more cost-effective for developers to run complex AI jobs anywhere, anytime.

Overview: CSGHub is an open-source platform designed for managing Large Language Models (LLM) assets, offering an efficient way to handle the entire LLM and their assets such as datasets, spaces and codes etc. It allows users to upload, download, store, verify, and distribute LLM assets like DeepSeek, Llama etc. via a web interface, `git` command line, a natural language Chatbot, or [CSGHub SDK](https://github.com/OpenCSGs/csghub-sdk). Key features: - Unified LLMs Management - Extensible Development Framework - Advanced Model Capabilities and Optimization - Space and Asset Management Assistant (Copilot) - Multi-Source Data Synchronization and Recommendation - Enterprise-Level Security and Access Control - On-Premises Deployment Solutions - E2E Data Processing and Intelligent Annotation System - Resilient High-Availability Architecture Tech stack: Vue Why it matters: CSGHub provides a user-friendly management platform for Large Language Models, enabling secure, efficient, and scalable operations. It offers a seamless integration with existing systems, making it ideal for enterprises looking to manage their LLM assets effectively. Repository: OpenCSGs/csghub GitHub stars: 4068 GitHub description: CSGHub is an open-source platform designed for managing Large Language Models (LLM) assets, offering an efficient way to handle the entire LLM and their assets such as datasets, spaces and codes etc. It allows users to upload, download, store, verify, and distribute LLM

Overview: oneDNN is an open-source library that provides optimized deep learning operations for CPUs and GPUs, serving as an acceleration layer for popular AI frameworks like PyTorch, TensorFlow, and OpenVINO. Key features: - Supports matrix multiplication, convolution, attention, normalization, and other operations. - Optimized for Intel 64/AMD64, AArch64, RISC-V, and Intel GPUs. - Hardware specialization at runtime, generating specialized kernels based on detected features. - Low-precision inference with support for `f32`, `bf16`, `f16`, `f8`, `f4`, `u8`, `s8`, and `u4`, `s4` data types. - Operation fusion for post-ops, including fused matmul and convolution. - Interoperability with various frameworks and applications through integration with OpenMP, TBB, SYCL, and OpenCL. Tech stack: C++ Why it matters: oneDNN accelerates deep learning computations, making it essential for optimizing performance in AI applications across different hardware platforms.

### Overview: Deepchecks is a comprehensive open-source solution designed to validate and continuously monitor AI and ML models and data. It provides a holistic approach from research to production, ensuring thorough testing and efficient collaboration for model deployment. ### Key Features: - **Deepchecks Testing**: Runs built-in and custom checks for tabular, NLP, and CV data validation. - **CI & Testing Management**: Collaborates on test results and iterates efficiently to ensure models are production-ready. - **Deepchecks Monitoring**: Tracks and validates deployed models' behavior in production. ### Tech Stack: - Python ### Why It Matters: Deepchecks offers a unified solution for AI and ML validation, enabling thorough testing and efficient collaboration throughout the development lifecycle. This makes it an essential tool for ensuring model reliability and data integrity in production environments.

Overview: PySR is an open-source tool designed for Symbolic Regression, a machine learning task aimed at finding interpretable symbolic expressions that optimize some objective. It leverages the Julia library SymbolicRegression.jl, providing high-performance, configurability, and ease of use. Key features: - Symbolic Regression - Symbolic Distillation (for Neural Networks) - Installation via pip or conda - AI Agent Skill integration Tech stack: Python, Julia Why it matters: PySR offers an efficient and powerful way to interpret deep neural networks, making it valuable for researchers and practitioners in various fields.

Overview: This project provides a tool to generate model summaries similar to Keras' `model.summary()` in PyTorch, making it easier to visualize and understand the architecture and parameters of deep learning models. Key features: - Model summary generation - Supports various PyTorch models - Compatible with both CPU and GPU - Provides detailed parameter counts and memory usage Tech stack: Python Why it matters: This tool simplifies the process of understanding complex neural network architectures, making it easier for developers to debug and optimize their models.

### Overview: Chinese-Novelist-Skill 是一个由开发者打造的AI写小说工具，旨在帮助用户快速生成完整且高质量的中文小说。通过三层问答、偏好记忆、自动校验和多种写作模式，该技能显著提升了创作效率和作品质量。 ### Key features: - **智能问答**：三层递进式问答系统，支持快速跳过和随机生成。 - **偏好记忆**：跨会话学习用户的喜好，下次创作更贴心。 - **中断续写**：意外中断时自动检测并从断点继续。 - **多种写作模式**：串行、子Agent并行、Agent Teams，按需选择。 - **自动校验**：完成后自动检查字数和连贯性，不合格自动重写。 - **每章必爽**：开头即高潮，结尾留悬念。 ### Tech stack: Python ### Why it matters: 写小说最难的是**坚持写完**。这个技能专为解决这个痛点而生，通过智能问答、偏好记忆、自动校验等技术，大大提高了创作效率和作品质量。无论是长篇网文连载还是短篇小说，都能轻松应对。

### Overview: This repository contains the PyTorch code for the paper "Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?" by Kensho Hara, Hirokatsu Kataoka, and Yutaka Satoh. It includes training, fine-tuning, and testing on Kinetics, Moments in Time, ActivityNet, UCF-101, and HMDB-51 datasets. ### Key Features: - **Model Training**: The repository includes training scripts for various ResNet architectures (ResNet-18, ResNet-34, ResNet-50) on different datasets. - **Fine-Tuning**: Scripts for fine-tuning models on specific datasets. - **Pre-trained Models**: Available pre-trained models for different datasets, including Kinetics-700, Moments in Time, and combined datasets. - **Distributed Training**: Support for distributed training using PyTorch. ### Tech Stack: - PyTorch ### Why It Matters: This repository provides a comprehensive set of tools for researchers and developers interested in action recognition using spatiotemporal 3D CNNs. The availability of pre-trained models and detailed scripts makes it an invaluable resource for both academic research and practical applications.