Deep Learning

PaddlePaddle is a comprehensive industrial deep learning platform that provides core frameworks, model libraries, and end-to-end development tools. It supports advanced features like unified dynamic and static graphs, automatic parallelism, and high-order differentiation for scientific computing. The platform is designed to facilitate large-scale model training and inference across diverse industrial sectors.

DeepGEMM is a unified CUDA library providing high-performance tensor core kernels specifically optimized for modern large language models. It features a lightweight Just-In-Time compilation module that eliminates the need for CUDA compilation during installation. The library delivers expert-tuned performance for various matrix operations, including FP8, FP4, and BF16 GEMMs, as well as fused MoE and MQA scoring.

PaddleFormers is a Transformers library built on the Baidu PaddlePaddle framework, designed to provide training interfaces and functional experiences for Large Language Models and Vision-Language Models equivalent to Hugging Face. By integrating tensor parallelism, pipeline parallelism, and automatic mixed precision, the project achieves training performance that surpasses Megatron-LM on mainstream models. Furthermore, it fully supports domestic computing chips and is compatible with the Safetensors format, helping developers efficiently complete the entire process from pre-training to post-training.

ROLL is an efficient, user-friendly library designed for scaling reinforcement learning workflows for large language models across large-scale GPU clusters. It supports diverse training paradigms including RLVR, agentic interaction, and distillation, while integrating advanced backends like Megatron-Core, vLLM, and SGLang. The framework provides robust observability and flexible resource management to enhance performance in complex reasoning and human preference alignment tasks.

RTP-LLM is a high-performance LLM inference acceleration engine developed by the Alibaba Foundation Model Inference team. This engine has been widely applied in various Alibaba business scenarios such as Taobao and Tmall, supporting multiple mainstream model formats and hardware backends. It provides efficient production-level services for large language models by integrating advanced operator optimization, quantization techniques, and distributed inference capabilities.

Index-AniSora is a powerful open-source framework designed specifically for high-quality anime video generation and animation production. The system features a comprehensive data processing pipeline, a controllable generation model with spatiotemporal masking, and a specialized evaluation benchmark. It supports diverse creative tasks including character 3D generation, video style transfer, and multimodal guidance for precise motion control.

TorchEasyRec is a PyTorch-based framework designed for developing production-ready deep learning recommendation models. It supports a wide range of tasks including candidate generation, ranking, multi-task learning, and generative recommendation. The framework offers high scalability, flexible data source integration, and seamless deployment options for real-world production environments.

PaddleCustomDevice is the custom hardware integration solution provided by the PaddlePaddle framework. Through standardized interface design, this project enables developers to integrate various third-party hardware backends into the PaddlePaddle ecosystem. It currently covers support for mainstream hardware platforms including Ascend, Cambricon, Intel GPU, and Apple MPS.

JaQMC is a modular, JAX-based framework designed for performing neural network quantum Monte Carlo simulations. It utilizes deep neural networks as variational wavefunctions to solve the electronic Schrödinger equation without relying on traditional basis sets. The project supports various quantum systems, including molecules, solids, and fractional quantum Hall states, through a highly configurable and extensible architecture.

vLLM Kunlun is a community-maintained hardware plugin that enables the seamless execution of vLLM on Kunlun XPU hardware. It utilizes a hardware-pluggable interface to decouple the integration process, ensuring compatibility with a wide range of open-source models. The project supports various architectures including Transformer-based, Mixture-of-Expert, and multi-modal LLMs on the Kunlun3 P800 platform.