Awesome AI Agents › Generative Software Frameworks

FedML-AI/FedML

⭐ 4065 Python added to this list on 2025-04-19 repository created 2020-07-21

FEDML is a unified and scalable machine learning library designed for large-scale distributed training, model serving, and federated learning. It supports running AI jobs anywhere at any scale, including decentralized GPUs, multi-cloud environments, edge servers, and smartphones. The project is closely integrated with TensorOpera AI, a generative AI platform that provides a comprehensive AI infrastructure with three interconnected layers: user-friendly MLOps, a well-managed scheduler, and high-performance ML libraries. TensorOpera AI enables developers to launch complex model training, deployment, and federated learning jobs easily, economically, and securely across various GPU clouds and on-premise clusters. The MLOps layer includes TensorOpera Studio, which allows access to popular open-source foundational models like large language models (LLMs), fine-tuning with custom data, and scalable deployment using TensorOpera Launch on a GPU marketplace. The Job Store maintains a collection of pre-built jobs for training, deployment, and federated learning, encouraging developers to run these jobs with customized datasets or models on cost-effective GPUs. The scheduler layer features TensorOpera Launch, which pairs AI jobs with the most economical GPU resources, automates provisioning, and manages job execution across different cluster topologies and configurations. It supports compute-intensive tasks such as large-scale training, serverless deployments, and vector database searches, and facilitates on-premise cluster management and hybrid cloud deployments. The compute layer includes TensorOpera Deploy for scalable, low-latency model serving; TensorOpera Train for distributed training of large foundational models; and TensorOpera Federate, a federated learning platform supporting on-device training on smartphones and cross-cloud GPU servers. The open-source FedML library underpins these capabilities, providing a unified and scalable foundation for running AI jobs at any scale. Overall, FEDML and TensorOpera AI together offer a powerful ecosystem for managing the entire lifecycle of AI workloads, from training and deployment to federated learning, across diverse and distributed computing environments.

https://github.com/FedML-AI/FedML

ai-agentai-jobsdeep-learningdistributed-trainingedge-aiedge-serversfederated-learningfederated-learning-platformgenerative-aigpu-cloudgpu-resourcesinference-enginelarge-scale-trainingmachine-learningmlopsmodel-deploymentmodel-servingmulti-cloudon-device-trainingon-premise-clustersopen-sourceschedulerserverless-deploymentsmartphonestensoropera-aivector-database