Awesome GitHub RepositoriesMemory Mapping Utilities

LLM101n is an educational machine learning curriculum and open-source resource designed to teach the fundamental principles and practical implementation of large language models. It functions as a technical manual that guides users through the end-to-end process of building and training neural network architectures from scratch using a dynamic tensor library for automatic differentiation and GPU-accelerated computation. The project distinguishes itself through interactive, notebook-based instruction that allows for real-time visualization of training processes. It supports rapid experimentati

This project is a low-dependency engine designed for training large language models using native C and CUDA. It provides a bare-metal environment for tensor computation, allowing for the execution of neural network operations directly on hardware accelerators without the overhead of high-level software abstractions. The framework distinguishes itself by implementing manual gradient backpropagation and custom hardware-specific kernels, providing granular control over memory mapping and computational precision. It supports distributed training across multiple graphics processors and compute nod

RPCS3 is a C++ application that functions as a software environment for executing and managing PlayStation 3 console titles on desktop operating systems. It operates by translating proprietary console hardware instructions and graphics commands into formats compatible with modern computer hardware, allowing for the execution of original game software. The emulator distinguishes itself through a comprehensive suite of tools for managing game libraries, applying software patches to modify performance characteristics, and tracking the compatibility status of individual titles. It includes integr

This project provides an open-source firmware platform and complete build environment for portable multi-tool hardware. It functions as an embedded operating system designed to manage radio, infrared, and physical interface components, enabling users to develop custom applications and system logic for specialized hardware devices. The firmware distinguishes itself through a modular architecture that organizes system functionality into isolated units, allowing for the development of custom user interfaces and logic. It includes a comprehensive collection of low-level drivers and applications s

This project is an educational resource for developing bare-metal operating systems and kernels from scratch on Raspberry Pi hardware. It provides a structured guide to systems programming using the Rust language, focusing on the implementation of core kernel components that execute directly on ARM-based hardware without the support of an underlying operating system or standard library. The tutorials emphasize a modular architecture that separates hardware-independent kernel logic from processor-specific and board-specific configurations. By utilizing a hardware abstraction layer and distinct

MNN is a high-performance inference engine and framework designed for on-device machine learning. It provides a comprehensive environment for executing, optimizing, and deploying neural network models directly on mobile and resource-constrained edge devices. The framework distinguishes itself through a robust model optimization toolkit that supports quantization, compression, and structural graph manipulation to minimize memory footprint and maximize execution speed. It features a modular architecture that abstracts hardware-specific backends, allowing models to run efficiently across diverse

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

Cutlass is a collection of C++ templates and Python interfaces for implementing high-performance linear algebra operations on NVIDIA GPUs. It provides a kernel composition framework for designing custom GPU kernels and a mixed-precision tensor library capable of executing operations across diverse data formats, ranging from 64-bit floating point to 4-bit integers. The project features a toolkit for operator fusion that integrates activation functions and bias calculations directly into matrix multiplication kernels to reduce memory passes. It also includes a Python-based domain-specific langu

Einops is a tensor manipulation library that provides a framework-agnostic interface for reshaping, Einstein summation, and multi-dimensional array operations. It serves as an abstraction layer that works across NumPy, PyTorch, TensorFlow, and JAX, allowing for tensor transformations without changing the API. The library distinguishes itself through a declarative notation system that uses readable string patterns to describe tensor rearrangements and reductions. This approach includes an extended Einstein summation interface that supports multi-letter axis names and a named dimension mapping

Torch7 is a scientific computing environment and tensor computation library used for deep learning research and numerical analysis. It functions as a Lua-based framework for training neural networks and learning agents, providing a toolkit for implementing architectures and training through reinforcement learning algorithms. The project is distinguished by its tight integration with C, utilizing a binding layer to map high-level scripting to low-level C structures for direct memory access. It supports hardware-accelerated computation by offloading linear algebra and convolution operations to

This project is a game AI training framework designed to develop and monitor reinforcement learning agents within a legacy game environment. It functions as a training and monitoring system that optimizes autonomous agents to complete game objectives through exploration and reward-based learning. The framework includes tools for game memory mapping and real-time trajectory visualization. These capabilities translate raw game memory addresses into visual coordinates, allowing agent movements and session data to be streamed to a map for the analysis of navigation patterns and area exploration.

This project is a Nintendo Entertainment System emulator written in Go. It functions as a hardware simulation that executes game ROMs by mimicking the original console's circuitry, including the processor and picture processing unit. The emulator includes a game library browser that identifies local files and retrieves titles and thumbnails from online databases using file checksums. It also implements a system of memory management mappers to ensure compatibility across various game ROM formats and hardware configurations. The software covers high-level capabilities for game ROM execution an

TileLang is a Python-embedded domain-specific language compiler that JIT-compiles and autotunes GPU kernels. It uses a tile-based DSL, automatic software pipelining, and parallel autotuning to generate optimized GPU kernels at runtime. It supports tensor core operations with Pythonic syntax, automatic memory management, and thread mapping. The compiler searches over tile sizes, thread counts, and scheduling policies, compiling and benchmarking candidates in parallel to find the fastest kernel. It also caches compiled binaries and tuning results to disk for reuse across sessions. TileLang inc

tiny-llm ist eine Inferenz-Engine für große Sprachmodelle und eine Transformer-Modell-Implementierung. Sie dient als Laufzeitumgebung für quantisierte Modelle und als Paged-Key-Value-Cache-Manager und bietet einen spezialisierten Inferenz-Stack, der für Apple Silicon optimiert ist. Das System zeichnet sich durch High-Throughput-Ausführungstechniken aus, einschließlich Continuous Batching und Paged Attention. Es nutzt ein Paged-Memory-System, um Fragmentierung während der Token-Generierung zu eliminieren, und verwendet On-the-Fly-Dequantisierung komprimierter Gewichte, um den Speicherbedarf während der Matrixmultiplikation zu reduzieren. Das Projekt deckt ein breites Spektrum an Modellarchitektur- und Performance-Funktionen ab, wie Mixture-of-Experts-Routing, Grouped Query Attention und Flash Attention. Es umfasst Unterstützung für fortgeschrittene Decoding-Logik, einschließlich Greedy Decoding und Sampling via Temperature, Top-K- und Top-P-Methoden. Die Implementierung ist in Python geschrieben und enthält benutzerdefinierte Low-Level-Kernel zur Beschleunigung der Tensor-Verarbeitung auf der Hardware.

Neural Prophet ist eine auf PyTorch basierende Bibliothek für Zeitreihenprognosen, die für interpretierbares Machine Learning entwickelt wurde. Sie dient als Dekompositions-Framework, das Signale in Bestandteile wie autoregressive Effekte, stückweise lineare Trends und Fourier-basierte Saisonalität zerlegt, um zukünftige Werte vorherzusagen. Das Projekt zeichnet sich durch die Kombination neuronaler Netze mit traditionellen Algorithmen aus, um Prognosen zu erstellen, die zugrunde liegende Trendtreiber erklären. Es bietet einen globalen Zeitreihen-Modellierungsansatz, der es ermöglicht, ein einzelnes Modell über mehrere gleichzeitige Reihen hinweg zu trainieren, um gelernte Muster zu teilen und gleichzeitig lokale Spezifitäten beizubehalten. Zudem fungiert es als Tool zur Unsicherheitsquantifizierung und nutzt Quantil-Regression und konforme Vorhersagen, um zuverlässige Prognoseintervalle zu generieren. Die Bibliothek bietet eine umfassende Suite an Funktionen für das Datenmanagement, einschließlich Abruf von Feiertagen, Lückenfüllung und Normalisierung. Sie deckt den gesamten Modellierungslebenszyklus mit automatisierter Hyperparameter-Optimierung, Erkennung von Trend-Changepoints und der Integration von zukünftigen sowie verzögerten Regressoren ab. Die Analyse wird durch Prognosedekomposition und Input-Attribution unterstützt, um zu visualisieren, wie spezifische Faktoren die finalen Vorhersagen beeinflussen.

Xarray is a Python multidimensional array library and labeled dataset framework. It extends the NumPy data structure by adding labels to arrays, allowing for the organization of complex N-dimensional data using named dimensions and coordinates. The library provides a NetCDF data interface for reading and writing scientific data formats such as NetCDF and Zarr. It enables scientific array computing by maintaining the relationship between data and physical coordinates during mathematical operations. The project covers multidimensional data analysis, geospatial data manipulation, and climate da

NuttX ist ein POSIX-konformes Echtzeitbetriebssystem, das für ressourcenbeschränkte eingebettete Umgebungen entwickelt wurde. Es fungiert als skalierbares Mikrocontroller-Betriebssystem, das eine Unix-ähnliche Umgebung für die Verwaltung von Hardware und die Ausführung von Anwendungen über Architekturen von 8-Bit bis 64-Bit hinweg bietet. Das System stellt eine hohe Software-Portabilität sicher, indem es einen Kernel implementiert, der POSIX- und ANSI-Standards folgt. Dies ermöglicht es Entwicklern, portable eingebettete Anwendungen unter Verwendung standardisierter API-Aufrufe über diverse Hardware-Architekturen hinweg zu erstellen. Das Projekt umfasst eine modulare Kernel-Architektur und eine Hardware-Abstraktionsschicht, um das System von spezifischen Chip-Peripheriegeräten zu entkoppeln. Es nutzt prioritätsbasiertes präemptives Scheduling für deterministische Antworten und bietet Tools zur Simulation von Hardwareumgebungen für das Testen von Firmware ohne physische Boards.

Rocket-chip is a framework for the parametric design, synthesis, and verification of RISC-V based processors and system-on-chip hardware. It functions as a generator that converts high-level specifications into synthesizable Verilog files for FPGA or ASIC implementation. The project utilizes a Scala-based hardware description framework to produce customizable pipelined processor cores, memory hierarchies, and peripheral devices. It employs a parameter-driven model and a two-phase negotiation process to resolve hardware interface specifications between modules during the elaboration phase. Th