Projects sharing features with Flash Attention

This project is an educational platform and research toolkit designed to teach deep learning through a combination of mathematical theory, visual diagrams, and executable code. It provides a comprehensive environment for building, training, and evaluating neural networks, grounding complex concepts in interactive computational notebooks that allow for hands-on experimentation. The framework distinguishes itself by interleaving theoretical foundations—including linear algebra, calculus, and probability—with practical implementations across multiple industry-standard libraries. It supports flex

DeepSpeed is a high-performance library designed to scale deep learning model training and inference across massive clusters of GPUs and compute nodes. It provides a comprehensive suite of tools for distributed training, enabling the execution of models that exceed the memory capacity of single devices through advanced parameter partitioning, pipeline-based model parallelism, and memory-efficient state offloading. The framework distinguishes itself through specialized communication-efficient optimizers and hardware-aware acceleration techniques. By utilizing gradient compression, quantization

Unsloth is a high-performance training and inference platform designed to optimize the lifecycle of large language and multimodal models. It provides a comprehensive engine for fine-tuning, executing, and managing models locally, with a focus on reducing memory consumption and increasing compute speed on consumer-grade hardware. The platform distinguishes itself through hand-optimized kernels and automated computational graph techniques that maximize hardware throughput. It supports advanced training methodologies, including reinforcement learning for reasoning and efficient adapter-based fin

This project serves as a comprehensive educational resource and technical guide for mastering deep learning through the PyTorch framework. It provides structured tutorials and practical code examples designed to teach core machine learning principles, ranging from fundamental tensor operations to the construction of complex neural network architectures. The repository distinguishes itself by bridging the gap between theoretical concepts and hands-on implementation. It covers the development of generative applications, such as image synthesis and style transfer, while offering guidance on opti

TensorRT-LLM is a platform and toolkit designed for compiling, optimizing, and serving transformer-based models on accelerated hardware. It functions as a framework that transforms machine learning models into efficient execution graphs, providing an engine to refine these models for specific hardware to maximize throughput and minimize latency during text generation. The project distinguishes itself through advanced execution strategies that manage the entire inference pipeline. It utilizes kernel-level fusion and static graph execution to optimize mathematical operations and computational f

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

jetson-inference is a set of libraries and tools for executing optimized deep learning models on embedded GPU hardware. Its primary purpose is to enable real-time computer vision and AI inference at the edge with low latency and high throughput. The project distinguishes itself through high-performance streaming analytics and the ability to execute concurrent AI pipelines on auto-grade silicon. It provides specialized support for multi-sensor stream processing, utilizing zero-copy data transport to load camera frames directly into GPU memory. The codebase covers a broad surface of capabiliti

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

Ludwig is a multimodal machine learning platform and low-code framework designed for building, training, and deploying neural networks. It enables the construction of models that process text, images, audio, and tabular data through a unified interface using declarative configuration files rather than custom code. The system features a specialized low-code framework for large language models, supporting supervised fine-tuning, preference alignment, and a constrained decoding tool to force structured data output via logit extraction. It also includes an automated model architecture search to i

Liger-Kernel is a collection of pre-built fused Triton kernels and patching utilities designed to accelerate large language model training. It provides drop-in kernel replacements for common LLM operations such as RMSNorm, cross-entropy loss, and attention, enabling increased throughput and reduced memory usage while preserving bitwise-exact gradients. The project serves as a toolkit for composing custom model architectures from individual optimized kernels and for patching pre-existing models with minimal code changes. The project distinguishes itself through its ability to perform runtime m

Fairseq is a PyTorch toolkit for sequence-to-sequence modeling, specializing in neural machine translation, automatic speech recognition, and large-scale language model training. It provides a framework for processing and aligning diverse data sources, including text, audio, and video, to support tasks such as speech-to-text conversion and multimodal sequence learning. The project is distinguished by its distributed training capabilities, which utilize parameter sharding, mixed-precision training, and CPU offloading to handle models that exceed single-device memory. It also includes specializ

This project is a collection of scripts and workflows for training, fine-tuning, and deploying large language models using the Hugging Face Transformers toolkit. It functions as a distributed training framework, a library for natural language processing task implementations, and a system for building retrieval-augmented generation chatbots. The repository includes specialized tools for model optimization, such as a Bayesian hyperparameter optimizer for automatically tuning model settings. It provides implementations for scaling model training across multiple graphics processors using data par

Firefly is a training framework and inference engine for large language models. It functions as a toolkit for pre-training and fine-tuning various open-weight architectures, providing a system for model alignment and parameter-efficient fine-tuning. The project includes utilities for merging adapter weights back into base models to create standalone files. It also provides a model alignment toolkit to format training data according to specific prompt templates, ensuring conversational consistency across different models. The framework supports distributed model training and preference-based

FastVideo is a comprehensive system for accelerated video generation, serving as a video generation inference engine, a video diffusion training framework, and a modular pipeline orchestrator. It provides a distributed transformer optimizer and a distillation toolkit designed to reduce denoising steps and model complexity to increase frame rates. The project distinguishes itself through specialized acceleration techniques, including joint distillation and sparse attention training. It implements low-step video generation and weight quantization to FP8 or FP4 precision to increase throughput a

minGPT is a minimal implementation of the Transformer architecture designed for training and experimenting with language models. It functions as a neural network training framework and a text generation engine, providing the necessary tools to manage data loading, backpropagation, and parameter updates for custom deep learning models. The project is structured as an educational resource for understanding how transformer architectures function by building and training models from scratch. It utilizes a modular block architecture and transformer-based self-attention to process sequences, allowi

LLaMA-Factory is a comprehensive suite for dataset preparation, model fine-tuning, memory optimization, and standardized API deployment. It provides a unified platform for the supervised and reward-based fine-tuning of large language models and vision-language models. The framework includes a specialized toolkit for training vision-language models and a model serving interface that deploys trained models through high-performance APIs. It utilizes precision tuning and quantization techniques to reduce the hardware requirements and memory footprint of large models. The system covers data pipel

This project is a training pipeline and framework for developing Chinese language models based on the Llama 2 architecture. It functions as a distributed GPU trainer and dataset preprocessing toolkit designed for both the initial pre-training of baseline models and subsequent supervised fine-tuning. The system distinguishes itself through a specialized workflow for Chinese text, incorporating a data curation pipeline that uses similarity hashing for deduplication and a tokenization process that converts raw text into memory-mapped binary files for efficient disk access. It implements a superv

FinGPT is a suite of specialized financial tools and a framework for adapting large language models to the financial domain. It provides a set of pipelines for financial entity extraction, sentiment analysis, and retrieval-augmented generation to improve the accuracy of financial information systems. The project distinguishes itself through efficient training workflows, utilizing low-rank adaptation and quantized low-rank adaptation to fine-tune models on consumer-grade hardware. It employs market-labeled datasets and reinforcement learning that uses actual stock price movements as reward sig

ipex-llm is an acceleration library and inference engine designed to optimize the execution and finetuning of large language models on Intel GPUs and NPUs. It provides a HuggingFace compatible model backend and a dedicated quantization toolkit for converting model weights into low-bit precision formats. The project facilitates distributed inference by splitting large model workloads across multiple accelerators using pipeline and tensor parallelism. It enables the deployment of models on Intel Arc, Flex, and Max GPUs to increase throughput and reduce latency. The library covers a broad range

XGBoost is a distributed machine learning library for implementing scalable gradient boosting decision trees used for regression, classification, and ranking. It functions as a predictive model framework and a cross-language toolkit, providing a core implementation with native bindings for Python, R, Java, Scala, and C++. The system is designed as a GPU-accelerated library that utilizes CUDA and NCCL to speed up the training of decision tree ensembles. It operates as a distributed framework capable of scaling training and prediction across multi-node clusters and GPU environments to process m

Torchtune is a PyTorch-native library for fine-tuning, aligning, and quantizing large language models. It provides a config-driven system for instantiating components, orchestrating distributed training, and managing parameter-efficient fine-tuning with quantization support, all through YAML-based configurations and command-line overrides. The library distinguishes itself through its comprehensive post-training workflow orchestration, combining supervised fine-tuning, preference optimization (DPO, PPO, GRPO), knowledge distillation, and quantization-aware training in a single configurable pip

nlp.js is a JavaScript natural language processing library and development framework used to build natural language understanding engines. It provides a toolkit for creating local machine learning models for intent classification and acts as a multilingual text processor that detects languages and normalizes text across various dialects. The framework distinguishes itself by supporting local execution on both servers and mobile devices, enabling chatbot functionality without an internet connection. It features a specialized system for conversational slot filling to collect mandatory informati

This is a structured deep learning curriculum for programmers, delivered as a collection of Jupyter notebooks. It teaches the fundamentals of training neural networks for computer vision, natural language processing, tabular data analysis, and collaborative filtering using PyTorch and the fastai library. The course is designed to be hands-on, guiding learners from building a training loop from scratch to fine-tuning pretrained models for a variety of practical tasks. The curriculum distinguishes itself by covering the full lifecycle of a deep learning project, from data preparation and augmen

AISystem is a comprehensive AI full-stack infrastructure project covering the entire pipeline from AI chip architecture to high-level training frameworks. It encompasses the development of AI compiler frameworks, inference engines, and distributed training orchestrators designed to coordinate workloads across a heterogeneous compute stack of CPUs, GPUs, and NPUs. The project focuses on the deep integration of software and hardware, employing software-hardware co-design to align tensor layouts with physical memory structures. It provides specialized capabilities for accelerating Transformer mo

This is a PyTorch deep learning implementation for training transformer-based language models. It functions as a distributed GPU trainer and framework designed to optimize text prediction models for increased speed and sample efficiency. The project is distinguished by its use of the Newton-Schulz weight optimizer. This method applies an iterative process to maintain semi-orthogonal parameter updates and weight matrices, which improves sample efficiency and reduces memory overhead during the training process. The framework covers broad capabilities in distributed GPU computing, including dat