ModelScope

ModelScope is an AI model hub with a Python library for developers and researchers who want to run models locally or adapt them to their own data. It covers language, vision, speech, multimodal tasks and scientific computing.

• Model discovery: Browse and download models such as Qwen, DeepSeek, Meta-Llama-3-8B-Instruct and LaMa image inpainting, alongside datasets for training and evaluation.
• Local inference: Use a shared interface to apply models to text, images, audio and video, including background removal, speech recognition and text generation.
• Fine-tuning and evaluation: Train models on your own datasets and assess their performance, with support for distributed training of large models.
• Application development: Customize inference and training components, then export and deploy models for use in applications.

The library is open source under Apache 2.0 and runs in a local Python environment, with Docker support. It works with PyTorch, TensorFlow and ONNX; individual models have their own framework requirements. Its hub connections handle model and dataset lookup, version management and caching. Most hub models are public and available for direct download.

Online model demos and ModelScope Notebook run on hosted infrastructure. The notebooks provide cloud CPU or GPU development environments, separate from the library you run on your own hardware.