Awesome GitHub RepositoriesSequence Parallelism Frameworks

DeepSpeed is a distributed deep learning optimization library and framework designed for the training and inference of massive AI models. It serves as a model parallelism orchestrator and a toolkit for scaling large language models across multiple GPUs and compute nodes. The project distinguishes itself through 3D parallelism orchestration, which combines data, pipeline, and tensor parallelism. It utilizes ZeRO-based memory partitioning to eliminate redundant storage and employs CPU-offload memory management to move weights and optimizer states to system RAM. Additionally, it provides special

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

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

This project is a distributed training infrastructure designed for aligning large language models through reinforcement learning. It functions as an end-to-end engine for complex alignment tasks, including proximal policy optimization, direct preference optimization, and iterative self-play. By providing a unified framework for multi-turn interactions and tool-use scenarios, it enables the development of models capable of reasoning and external environment engagement. The framework distinguishes itself through a decoupled architecture that separates model training from sample generation. This

Megatron-LM is a distributed transformer training library and large language model training framework designed to scale models across thousands of GPUs. It functions as a GPU-optimized deep learning toolkit and a scaling engine for mixture-of-experts architectures, enabling the training of models with hundreds of billions of parameters. The project implements multi-dimensional model parallelism, combining tensor, pipeline, data, expert, and context-based workload distribution. It specifically optimizes mixture-of-experts architectures through integrated memory and communication improvements t

HunyuanVideo is a generative artificial intelligence framework designed to synthesize high-fidelity video sequences from descriptive text prompts. It utilizes a latent diffusion architecture that compresses video data into compact representations, allowing for the generation of dynamic visual content while maintaining temporal and spatial fidelity. The system distinguishes itself through a specialized inference engine that supports eight-bit weight quantization and sequence-parallel distribution. These capabilities enable the execution of large-scale generative models on hardware with limited

Open-Sora-Plan is a text-to-video framework and distributed video training system. It utilizes a diffusion transformer architecture and large language model components to transform written descriptions or image prompts into high-quality video sequences. The system features a distributed infrastructure designed for large-scale video training and inference. It employs sequence parallelism to split high-resolution or long-duration video samples across multiple GPUs and uses a sparse attention mechanism to increase processing speed. The project includes capabilities for both text-to-video and im

OpenRLHF is a training framework and alignment library designed for reinforcement learning from human feedback across distributed GPU clusters. It provides tools for aligning large language models and multimodal vision-language models using algorithms such as PPO, GRPO, and DPO. The framework distinguishes itself through a distributed inference engine that overlaps sample rollout with training to increase throughput. It supports scaling to models exceeding 70 billion parameters via parameter sharding and handles long-context sequences through ring-attention sequence parallelism. The project

This project provides a comprehensive technical guide and framework for engineering large-scale machine learning systems. It covers the full lifecycle of model development, focusing on the infrastructure and computational principles required to build, train, and serve generative AI models across distributed GPU clusters. The repository distinguishes itself by offering deep-dive tutorials and implementation strategies for complex system challenges. It emphasizes high-performance architectural primitives, such as collective communication orchestration, distributed tensor sharding, and static gr

xtuner es un motor de entrenamiento integral para modelos de lenguaje grandes, que ofrece un toolkit para pre-entrenamiento, ajuste fino supervisado (fine-tuning) y la optimización de modelos multimodales de visión-lenguaje. Sirve como un acelerador de entrenamiento distribuido y un framework especializado para escalar modelos de Mezcla de Expertos (MoE) y alinear el comportamiento del modelo mediante aprendizaje por refuerzo a partir de retroalimentación humana (RLHF). El proyecto se distingue por optimizaciones avanzadas de memoria y cómputo, como el paralelismo de secuencia para ventanas de contexto ultra largas y el paralelismo de pipeline entrelazado para reducir el tiempo de inactividad de la GPU. Proporciona una suite dedicada para la optimización de preferencias, implementando técnicas como Group Relative Policy Optimization y Direct Preference Optimization para refinar las políticas del modelo y los sistemas de recompensa. Las áreas de capacidad cubren el entrenamiento distribuido de modelos a través de múltiples nodos, la preparación de datasets multimodales y la gestión del ajuste fino basado en adaptadores. El motor también incluye herramientas para la evaluación de modelos, fusión de pesos y exportación de parámetros entrenados a motores de inferencia. El entrenamiento se gestiona mediante archivos de configuración estandarizados y lanzadores distribuidos para asegurar resultados consistentes a través de clusters de computación.