Awesome GitHub RepositoriesComponent Registries

This project is a modular research toolkit designed for developing, training, and evaluating deep learning models for object detection, segmentation, and video instance tracking. It provides a flexible training engine that manages complex neural network execution, including distributed training, custom lifecycle hooks, and weight optimization. The framework is built around a hierarchical configuration system that allows users to define architectures, data pipelines, and training hyperparameters through composable, inheritable files. The project distinguishes itself through its highly modular

Fairseq is a deep learning research toolkit and sequence-to-sequence framework built on PyTorch. It provides a system for training and deploying models that map input sequences to output sequences, with a primary focus on neural machine translation and speech recognition. The toolkit allows for the generation of text sequences through search algorithms such as beam search and nucleus sampling. It includes capabilities for producing synthetic parallel training data by translating monolingual text using reverse sequence models. The framework supports large scale model training through multi-de

Folly is a collection of high-performance C++ components designed as an extension to the C++ Standard Library for large-scale production environments. It provides specialized toolkits for memory management, concurrency, asynchronous workflows, and low-latency input and output operations. The project distinguishes itself through the provision of lock-free containers and bounded queues to minimize contention in multi-threaded applications, alongside a framework for managing deferred computations using futures and promises. It further offers specialized memory arenas and optimized implementation

Catch2 is a comprehensive framework for C++ software validation, providing an environment for unit testing, integration verification, and performance analysis. It enables developers to define and execute automated test suites and micro-benchmarks directly within their applications. The framework is distinguished by its header-only distribution, which allows for integration into existing build systems without requiring complex external dependencies. It utilizes a hierarchical section-based execution model that supports behavior-driven testing, allowing for shared setup and teardown logic acros

Bit is a component-based development platform and monorepo orchestrator used to build, manage, and share reusable software components across projects. It functions as a system for modular software architecture, providing a component registry for publishing and distributing independent software modules via remote scopes. The platform distinguishes itself through lane-based versioning, which isolates feature development into parallel tracks for comparison and merging. It utilizes a scope-based namespace registry to organize components into hierarchical groups and employs environment-driven comp

Plate is a headless editor toolkit and framework designed for building custom rich text editing experiences. It provides a modular architecture that relies on a plugin-based system to manage document models, schema-driven node definitions, and state updates, allowing developers to assemble tailored content creation environments. The project distinguishes itself through a source-code injection pattern, where pre-built user interface templates and components are copied directly into the local project codebase. This approach ensures that developers maintain full control over the styling, markup,

This project is a web component framework and optimized web markup standard designed for high performance web development. It provides a system for building fast-loading websites using a specialized set of HTML components and scripts, complemented by a web performance validation suite to ensure markup compliance. The framework includes a dynamic HTML template engine for rendering data-driven content without full page reloads and a dedicated ad network integration framework. This integration system manages third-party advertisements with built-in viewability metrics and optimized loading seque

Puck is a visual page editor and layout tool for React. It functions as a CMS page builder and component orchestrator, allowing for the design and arrangement of structured content pages through a drag-and-drop interface. The system utilizes a pluggable component registry to integrate external React components into the visual canvas. It employs schema-driven mapping and JSON-based serialization to store and persist page structures as portable data objects. The platform covers the domain of no-code page building and CMS content management, providing a visual reconciliation system to synchroni

EnTT is a C++ library designed for data-oriented design and entity component system architecture. It provides a framework for managing game objects and simulation states by separating entity data from logic, allowing for the efficient organization and manipulation of large collections of related data objects. The library utilizes sparse sets to store entities and components in contiguous memory, which facilitates cache-friendly iteration and constant-time lookups. It employs template metaprogramming for compile-time type reflection and type-erasure techniques to provide a unified interface fo

AllenNLP is a PyTorch-based research library and deep learning language toolkit designed for developing and training neural network architectures for linguistic tasks. It provides a distributed training system that coordinates data and gradients across multiple GPUs and a framework for integrating pretrained transformer architectures. The system distinguishes itself with a dedicated algorithmic bias mitigation tool used to identify and reduce bias in linguistic model predictions. It also includes model influence analysis to interpret predictions by calculating the influence of specific traini

Cosmos is a UI component sandbox and development tool used for building and testing user interface elements in isolation from main application logic. It serves as a frontend design system tool and a modular prototyping environment for verifying the visual behavior and functionality of reusable interface components. The tool provides a dedicated workspace for iterating on visual elements and their states without requiring a full application deployment. This environment supports a frontend design system workflow by allowing components to be developed and documented independently before project

mmaction2 es un kit de herramientas de comprensión de video de PyTorch diseñado para entrenar y evaluar modelos de aprendizaje profundo. Sirve como un framework para el reconocimiento de acciones, localización temporal y detección de acciones espaciotemporales, proporcionando herramientas especializadas tanto para el análisis de video basado en píxeles como para el reconocimiento de acciones basado en esqueletos. El proyecto se distingue por una arquitectura modular que cuenta con descubrimiento de componentes basado en registro y ensamblaje de modelos jerárquico impulsado por configuración. Admite la fusión de características multimodales, integrando marcos RGB, flujo óptico y audio, e incluye capacidades para la recuperación de clips de video mediante texto y predicción de video zero-shot. A grandes rasgos, el framework cubre la ingeniería de conjuntos de datos de video, incluyendo la estandarización de anotaciones y el muestreo de fotogramas, así como el entrenamiento y evaluación integral de modelos. Proporciona utilidades para el entrenamiento distribuido, destilación de conocimientos y optimización de inferencia mediante reparametrización de modelos. La base de código admite la exportación de modelos ONNX y la contenerización del entorno para el despliegue a través de diferentes nodos de cómputo.

mmocr es un framework de reconocimiento óptico de caracteres basado en PyTorch diseñado para entrenar y desplegar modelos de detección de texto, reconocimiento y extracción de información clave. Sirve como una caja de herramientas integral para la detección y reconocimiento de texto en escenas, proporcionando bibliotecas especializadas para localizar regiones de texto y convertir texto visual en cadenas codificadas por máquina. El proyecto se distingue por un framework de investigación para la extracción de información clave y capacidades avanzadas de detección de texto. Estas incluyen la detección basada en puntos utilizando transformers y el uso de curvas de Bezier parametrizadas para identificar y transcribir texto con formas arbitrarias. El framework cubre una amplia superficie de capacidades de visión artificial, incluyendo la gestión de pipelines de datos para aumentar y estandarizar diversos conjuntos de datos OCR, entrenamiento de modelos con escalado distribuido y evaluación del rendimiento utilizando métricas OCR estándar. También proporciona utilidades para la manipulación de polígonos geométricos y visualización de resultados para auditar predicciones contra anotaciones de verdad fundamental. El sistema está implementado en Python y admite la instalación mediante empaquetado de entorno Docker.

AdalFlow es un framework de agentes de IA autónomos y una librería de aplicaciones LLM diseñada para construir flujos de trabajo modulares. Sirve como una interfaz agnóstica al modelo y orquestador de pipelines RAG, permitiendo a los usuarios desarrollar agentes ReAct que utilizan razonamiento iterativo y ejecución de herramientas externas para resolver tareas complejas. El proyecto se distingue por un sistema de optimización de prompts que utiliza descenso de gradiente textual para refinar automáticamente las plantillas de prompts y ejemplos de pocos disparos (few-shot). Trata la retroalimentación del modelo como una señal diferenciable, permitiendo una forma de retropropagación de LLM para mejorar iterativamente la calidad de la salida basada en métricas de evaluación. El framework cubre una amplia superficie de capacidades, incluyendo generación aumentada por recuperación (RAG) con búsqueda semántica vectorial y reranking, rastreo de ejecución basado en spans para observabilidad y análisis estructurado basado en esquemas. Proporciona una capa de comunicación unificada para numerosos proveedores de modelos propietarios y de código abierto, y admite la conversión de funciones de Python en interfaces de herramientas estandarizadas. El sistema está implementado en Python y se integra con MLflow para el seguimiento y análisis de flujos de trabajo.

mmtracking is a PyTorch video perception framework designed for training and deploying computer vision models that analyze sequential image data. It provides specialized tools for multi-object tracking, video instance segmentation, and a configuration-driven system for managing deep learning models. The project utilizes a deep learning model registry and a configuration-driven pipeline to swap model backbones and detectors without modifying the core codebase. This modular approach allows for the development of custom perception architectures by combining various components and configurations.

This project is a technical resource and pattern library for building enterprise applications with Python. It serves as a guide for implementing clean architecture, providing a framework for separating core business logic from infrastructure and external frameworks. The material focuses on Domain-Driven Design and the application of architectural patterns to maintain complex business requirements. It provides specific guidance on the Repository pattern for data abstraction, Command-Query Responsibility Segregation for optimizing read and write paths, and the use of dependency inversion to dec

Rest_rpc is a C++20 remote procedure call framework designed for building distributed services. It provides an asynchronous service interface that allows developers to expose local functions as network-accessible services, enabling remote clients to trigger tasks through automated serialization and communication protocols. The framework distinguishes itself by leveraging template metaprogramming to perform compile-time type introspection, which generates binary encoders and decoders without the need for runtime reflection. It further optimizes network throughput through zero-copy buffer manag