Awesome GitHub RepositoriesOnline Quantization

Sglang is a high-performance inference engine and serving system designed for large language and multimodal models. It provides a programmable interface for orchestrating complex generation workflows, enabling developers to coordinate multi-turn dialogues, tool invocations, and reasoning chains through a domain-specific language. The platform is built to support production-scale deployments, offering an OpenAI-compatible API that allows for integration with existing application ecosystems. The system distinguishes itself through a disaggregated architecture that separates compute-intensive pr

faster-whisper is an automatic speech recognition framework and an optimized implementation of the Whisper speech-to-text engine. It functions as a CTranslate2 inference engine designed to convert spoken audio into written text. The project serves as a model quantization tool that transforms large audio model weights into lower precision formats. This process reduces memory usage and increases execution speed on hardware by utilizing integer quantized weights. The framework covers a broad range of capabilities including batch audio transcription for parallel processing and voice activity det

TurboVec is a high-performance Rust vector database and quantized search index designed for storing and retrieving high-dimensional embeddings. It functions as a pluggable vector store for large language model orchestration frameworks, providing a memory-efficient alternative to standard in-memory storage. The project distinguishes itself through a high-dimensional vector compressor that utilizes random rotation and data-oblivious scalar quantization to reduce memory footprints. Retrieval is accelerated via SIMD kernels that process distance calculations and search operations for increased th

This is a PyTorch implementation of a text-to-image model designed for synthesizing high-fidelity images from natural language descriptions. It utilizes a diffusion image generator to transform latent embeddings into visual data through an iterative denoising process. The system employs a two-stage latent mapping process, using a CLIP-based latent prior to map text embeddings to image embeddings before decoding them into pixels. It features a cascading diffusion decoder that produces high-resolution imagery by passing low-resolution outputs through a sequence of models at increasing scales.

BasicSR is a PyTorch-based image restoration toolbox and framework designed for training and deploying deep learning models to upscale, denoise, and deblur images and videos. It serves as a comprehensive system for image super-resolution and video quality restoration, providing the necessary infrastructure to recover fine visual details and increase pixel density. The project distinguishes itself through specialized toolkits for facial image enhancement and high-fidelity face synthesis, as well as a dedicated video quality restoration suite that utilizes deformable convolutions and generative

This project is a comprehensive collection of educational examples and reference implementations for building vision and language models using PyTorch. It serves as a deep learning tutorial covering the end-to-end process of developing neural networks, from initial architecture definition to final production deployment. The repository provides detailed guides on implementing a wide range of domain-specific models, including convolutional neural networks for object detection and segmentation, as well as transformer and recurrent architectures for natural language processing. It emphasizes gene

ccv is a computer vision library written in C designed for high-performance visual analysis. It serves as a framework for image classification, object detection, and the identification of faces, pedestrians, and vehicles. The library distinguishes itself through hardware-accelerated vision and deep learning inference optimizations. It utilizes a quantized tensor processor to transform floating-point data into eight-bit integers and implements integer-quantized attention mechanisms to reduce memory bandwidth and increase data throughput. The project covers a broad range of capabilities, inclu

mistral.rs is an inference engine for large language models that runs locally and exposes models behind OpenAI and Anthropic-compatible APIs. It serves as a multi-model serving platform, capable of loading several models in a single server process with per-request routing and on-demand loading and unloading. The engine supports multimodal inference, processing text alongside images, video, audio, and speech inputs, and includes a quantized model deployment runtime that reduces memory use and speeds up inference on consumer hardware. The project distinguishes itself through an agentic tool exe

mmcv is a foundation library for computer vision based on PyTorch. It provides a comprehensive system for constructing convolutional neural networks, a toolkit for image and video preprocessing, and a collection of high-performance deep learning vision operators. The project is distinguished by its hardware-accelerated kernels for complex operations such as deformable convolutions and region pooling. It features a configuration-driven framework that allows for the dynamic instantiation of network layers and the registration of custom modules without modifying code. The library covers a broad

هذا المشروع عبارة عن مورد تعليمي شامل ودورة تدريبية لبناء الشبكات العصبية باستخدام PyTorch. يغطي اللبنات الأساسية للتعلم العميق، بما في ذلك معالجة الموترات (tensors)، والتمايز التلقائي، وبناء مكونات الشبكة العصبية المعيارية. يعمل المستودع كدليل تقني للعديد من المجالات المتخصصة. يوفر تفاصيل تنفيذ لمهام رؤية الكمبيوتر مثل تصنيف الصور، واكتشاف الكائنات، والتجزئة الدلالية، بالإضافة إلى سير عمل معالجة اللغات الطبيعية التي تتضمن المحولات (transformers)، والشبكات المتكررة، والنماذج التوليدية. بالإضافة إلى ذلك، يتضمن مرجعاً للذكاء الاصطناعي التوليدي، مع التركيز بشكل خاص على تركيب الصور عبر نماذج الانتشار (diffusion models) والشبكات التنافسية. تمتد المادة إلى تحسين النماذج وخطوط أنابيب النشر. تغطي تقنيات لتقليل حجم النموذج وزيادة سرعة الاستنتاج من خلال التكميم (quantization) وتصدير النماذج إلى تنسيقات مثل ONNX وTensorRT. تشمل مجالات القدرة الأخرى هندسة البيانات للتحميل المتوازي، وتقييم النموذج باستخدام مقاييس مخصصة، ونشر نماذج اللغات الكبيرة مفتوحة المصدر. يتم تقديم المشروع بشكل أساسي كسلسلة من دفاتر Jupyter.

Thumbhash هي مكتبة تشفير صور تحول الصور إلى سلاسل مدمجة لاستخدامها كعناصر نائبة خفيفة الوزن. توفر هذه الصيغة تمثيلاً ثنائياً لصورة يحافظ على اللون ونسبة العرض إلى الارتفاع، مما يسمح بتوليد وتصيير معاينات ضبابية أثناء تحميل الصفحة. يعمل المشروع كبديل لـ Blurhash من خلال توفير طريقة لتشفير الصور في سلاسل صغيرة مع تجنب العبء الحسابي عالي التردد. يعمل أيضاً كمستخرج لوحة ألوان، مما يتيح اشتقاق متوسط لون RGB المهيمن من تمثيل مشفر لاستخدامه في الخلفيات ذات اللون الصلب. تغطي المكتبة مجموعة من قدرات معالجة الصور، بما في ذلك استخراج اللون المهيمن، وتشفير الصور المدمج، وتصيير العناصر النائبة لدعم واجهات التحميل التكيفية. تستخدم هذه العمليات تحويلات جيب التمام المنفصلة والاستيفاء ثنائي الخطية لإعادة بناء تقريبات بصرية منخفضة الدقة من معاملات التردد المخزنة.

FastDeploy is a high-performance deployment framework for large language models, vision models, and multimodal models. It provides the infrastructure to launch model services that process combined image, video, and text inputs, exposing these capabilities through a standardized, OpenAI-compatible API for chat and text completions. The project distinguishes itself through advanced inference pipeline engineering and GPU optimization. It employs speculative decoding, tensor parallelism, and a disaggregated execution model that separates prefill and decode phases across different hardware resourc

PocketFlow is an integrated toolkit for deep learning model compression, distributed training, and mobile format optimization. It provides a system for reducing the size and complexity of neural networks to improve inference efficiency, featuring a dedicated engine for knowledge distillation and a mobile model optimizer. The framework differentiates itself through an automated hyperparameter tuning system that uses reinforcement learning and statistical models to determine optimal compression ratios and layer-wise bit allocation. It also includes a distributed training system that utilizes mu

llm-compressor is a quantization toolkit and post-training library designed to reduce the memory footprint and size of large language models. It provides a framework for compressing models using weight and activation quantization to enable more efficient deployment. The project distinguishes itself through a distributed quantization framework that utilizes data-parallel processing and disk-based weight offloading to handle massive model checkpoints that exceed available system memory. It includes specialized compressors for diverse architectures, including Mixture-of-Experts, Vision-Language,