Awesome GitHub RepositoriesModel Architecture and Evaluation

DeepSeek-V3 is a large language model that provides comprehensive resources for model utilization, including technical specifications, pre-trained weights, and evaluation benchmarks. The project details the core transformer architecture, including parameter counts and multi-token prediction modules, while supporting native 8-bit floating-point quantization. The repository offers extensive support for local and distributed inference through integration with multiple frameworks and engines. It includes documentation for deploying the model across various hardware configurations, such as GPUs an

This repository serves as a centralized collection of state-of-the-art deep learning architectures and reference implementations designed for research and application development. It provides a comprehensive toolkit for computer vision and natural language processing, offering pre-built models and training pipelines for tasks ranging from image classification and object detection to complex sequence modeling. The project distinguishes itself by providing a flexible execution harness that manages the entire training lifecycle, including data ingestion and backpropagation. It supports scalable

ONNX is an open-source standard for machine learning interoperability that provides a unified format for representing neural network models. By defining a common set of operators and a standardized file structure, it enables models to be shared, exported, and executed consistently across different training frameworks and software ecosystems. The project functions as an intermediate representation layer that decouples model development from deployment. It utilizes a language-neutral binary serialization format to store model structures and weights, ensuring that computational graphs remain por

OpenLLM is a framework for deploying, managing, and scaling open-source large language models

This project provides an animated Live2D character widget that can be embedded on any web page as an interactive mascot. The widget renders characters using the Cubism SDK on an HTML canvas, and can be deployed either via a content delivery network for zero-setup integration or self-hosted on a personal server for full control over asset delivery. The mascot responds to visitor actions through CSS selector-based interaction binding, displaying custom speech bubbles when users hover over or click specific page elements. Visitors can click and drag the character to reposition it anywhere on the

OpenVINO is an AI inference engine and model serving platform designed to execute optimized deep learning models across CPUs, GPUs, and NPUs through a unified API. It includes a model optimization toolkit for converting, quantizing, and compressing models from various frameworks, alongside a specialized generative AI runtime for large language models. The project distinguishes itself through a plugin-based hardware acceleration layer that maps neural network operations to vendor-specific drivers. It features advanced execution mechanisms such as continuous batching, speculative decoding, and

This repository provides a collection of reference implementations and code examples for training and deploying machine learning models using the MLX framework. It serves as a practical guide for executing distributed training, fine-tuning large language models, converting model weights, and implementing multimodal generative workflows. The project distinguishes itself through specialized examples for local hardware execution, featuring weight quantization to reduce memory usage and low-rank adaptation for parameter-efficient fine-tuning. It also includes scripts for transforming external mod

MochiDiffusion is a local client for Stable Diffusion that functions as an AI image generation studio. It provides a workspace for performing text-to-image, image-to-image, and inpainting tasks, enabling the production of high-resolution images offline using local hardware and neural engine acceleration. The project includes a local model manager for importing, organizing, and converting machine learning models into compatible formats for offline execution. It features a ControlNet integration tool to guide structural composition and spatial layout, alongside a dedicated image upscaler that u

StabilityMatrix is a centralized installer and orchestrator for Stable Diffusion web interfaces and their dependencies. It functions as a generative AI workspace and portable runtime, providing a unified interface to install and update AI image generation packages within isolated environments to prevent global system conflicts. The project distinguishes itself through a shared model manager that imports, organizes, and shares checkpoints across different installations. It utilizes a central model repository and filesystem mapping to allow multiple packages to access the same large binary asse

Pythia est un framework de recherche multimodal et un système d'entraînement distribué conçu pour construire, entraîner et évaluer de grands modèles qui combinent des données visuelles et linguistiques. Il fournit un environnement modulaire pour développer des modèles vision-langage, se concentrant sur l'intégration d'entrées d'images et de texte dans des représentations de caractéristiques partagées. Le framework utilise une architecture modulaire qui découple les blocs de construction du modèle en composants interchangeables, permettant une configuration flexible des modules de vision et de langage. Il inclut une suite de benchmarks pour exécuter des modèles de référence contre des jeux de données standardisés afin d'établir des bases de performance cohérentes pour les tâches vision-langage. Le système prend en charge des pipelines d'entraînement distribués pour mettre à l'échelle le développement de modèles sur plusieurs nœuds de calcul et utilise des fichiers de configuration externes pour le mappage des hyperparamètres afin d'assurer la reproductibilité de la recherche.

Ce dépôt fournit une collection de modèles d'apprentissage profond et d'architectures de réseaux de neurones construits pour les tâches de traitement du langage naturel. Il fonctionne comme une bibliothèque de modèles pré-entraînés conçus pour traiter, analyser et générer des données de langage humain en utilisant le framework TensorFlow. Le projet utilise la modélisation séquence-à-séquence et des architectures neuronales en couches pour gérer les données de langage de longueur variable. En employant des graphes de flux de données statiques et des représentations basées sur des tenseurs, les modèles exécutent des opérations mathématiques pour transformer les caractéristiques d'entrée en significations linguistiques abstraites. Les utilisateurs peuvent charger des sérialisations de poids pré-entraînés pour initialiser ces réseaux pour des tâches d'inférence spécifiques. La collection couvre un large éventail de capacités, y compris la classification de texte, la traduction linguistique et la reconnaissance d'entités nommées. Elle prend également en charge la transcription parole-vers-texte, la synthèse texte-vers-parole, le résumé de documents et la génération de contenu textuel conversationnel ou original. De plus, le dépôt inclut des outils pour analyser les structures grammaticales et visualiser les poids d'attention pour interpréter les modèles de décision des modèles.