✳ainav.
ModelScope screenshot
ModelScope website preview

PRODUCT OVERVIEW

What is ModelScope?

ModelScope is an open-source model community and innovation platform founded in June 2022 by the Institute for Intelligent Computing together with the CCF Open Source Development Committee. It hosts AI models, datasets, and Studios, and provides computation infrastructure for model serving. The ModelScope Library is a Python library for efficient model inference, fine-tuning, and evaluation across vision, speech, and LLM tasks. Related open-source frameworks include EvalScope for evaluation and benchmarking, Swift for LLM training and inference, and ModelScope-Agent for connecting models with the world. Studios offer a free and flexible space to build and demonstrate AI applications from model capabilities. It serves researchers, developers, and teams building AI applications. Sign-up grants 200 Magicubes, with 50 more daily for linked Alibaba Cloud accounts.

How to use ModelScope

  1. 1

    Create an account

    Sign up or log in on ModelScope to access models, datasets, Studios, and community features.

  2. 2

    Explore models and datasets

    Browse the Models and Datasets sections by task, such as computer vision, NLP, voice, multimodal, and scientific computing.

  3. 3

    Use the ModelScope Library

    Install the Python library to run model inference, fine-tuning, and evaluation across vision, speech, and LLM models.

  4. 4

    Build a Studio application

    Use the free Studio space to build and demonstrate AI applications based on model capabilities on the platform.

  5. 5

    Evaluate and train with frameworks

    Use EvalScope for benchmarking and Swift for LLM training and inference with supported models and methods such as LoRA.

ModelScope Features

Open-source hub for hosting AI models and datasets

ModelScope Library for inference, fine-tuning, and evaluation

Studios for building and demonstrating AI applications

EvalScope framework for model evaluation and benchmarking

Swift framework for LLM training and inference

ModelScope-Agent framework for connecting models with the world

Task coverage across vision, NLP, voice, multimodal, and scientific computing

Computation infrastructure for model serving

ModelScope Use Cases

  • Discover and download open-source AI models and datasets

  • Fine-tune and evaluate models with Python-based tooling

  • Build and showcase AI applications in Studios

  • Benchmark large models with EvalScope

  • Train and infer LLMs with Swift using methods such as LoRA

  • Connect ModelScope models to external tools with ModelScope-Agent

ModelScope FAQs

ModelScope is an open-source model community and innovation platform that hosts AI models, datasets, and Studios, and provides computation infrastructure for model serving.

It was founded in June 2022 by the Institute for Intelligent Computing together with the CCF Open Source Development Committee.

The Python library supports efficient model inference, fine-tuning, and evaluation for models ranging from vision and speech to LLMs.

Studios provide a free and flexible space to build and demonstrate AI applications based on the atomic capabilities of models on the platform.

EvalScope is a customizable framework for efficient large model evaluation and performance benchmarking.

Swift supports models such as LLaMA, Qwen, ChatGLM, and Baichuan, and training methods including LoRA, ReSTuning, and NEFTune.

It is an agent framework that connects ModelScope model capabilities with the world.

Signing up or logging in grants 200 Magicubes, and linking an Alibaba Cloud account grants 50 more daily.

ModelScope Official Information

Company
ModelScope
Website
https://modelscope.cn/
Support
contact@modelscope.cn
Documentation
https://modelscope.cn/docs

Contact ↗

ModelScope Pricing

Not specified

No consumer subscription prices are listed on the supplied pages. Sign-up grants 200 Magicubes, and linking an Alibaba Cloud account grants 50 more daily.

See the official website for current pricing.

COMMUNITY REVIEWS

ModelScope User Reviews

Reader reviews

No reviews yet.

Share your experience with ModelScope

Sign in to share your experience. Your display name, rating and review will be public after approval. One review per account per tool.

Your rating