ModelScopeBuilt on the concept of Models as a Service (MaaS). It aims to bring together state-of-the-art machine learning models from the AI ​​community and simplify the process of leveraging AI models in real-world applications. The open source core ModelScope library in this repository provides interfaces and implementations that allow developers to perform model inference, training, and evaluation.

In particular, with a rich API abstraction layer, the ModelScope library provides a unified experience to explore state-of-the-art models spanning CV, NLP, speech, multimodality, and scientific computing. Model contributors from different domains can integrate models into the ModelScope ecosystem through layered APIs, allowing easy and unified access to their models. Once integrated, model inference, fine-tuning, and evaluation can be done with just a few lines of code. At the same time, flexibility is provided so that different components in the model application can be customized if required.

In addition to containing implementations of various models, the ModelScope library also supports the necessary interaction with the ModelScope backend services, in particular with the Model-Hub and Dataset-Hub. This interaction facilitates the management of various entities (models and datasets) to perform seamlessly in the background, including entity lookup, version control, cache management, and more.

Hundreds of models in ModelScopePublicly available on (700+ and counting), covering the latest developments in NLP, CV, audio, multimodality, and AI for Science. Many of these models represent SOTA-specific domains in their models and are open sourced for the first time on ModelScope. Users can access the ModelScope (modelscope.cn), and through the online experience, you can experience the performance of these models for yourself with just a few clicks.Immediate developer experience is also available through ModelScope Notebook, powered by a cloud-ready CPU/GPU development environment – just one click on ModelScopecan.