无问芯穹

Introduction to Infini-AI

Infini-AI, also known as 无问芯穹 (Wú Wèn Xīn Qióng), is a cutting-edge AI model-as-a-service (MaaS) platform developed by Infinigence AI. It aims to provide efficient and unified deployment of large AI models across various chip architectures, including NVIDIA, AMD, and other domestic chips. The platform leverages industry-leading AI computing optimization capabilities and computing solutions to accelerate the implementation of large models in the era of Artificial General Intelligence (AGI).

Key Features

• Multi-Model Support: Infini-AI supports a wide range of AI models, including Llama-2, Qwen, ChatGLM, and Infini-Megrez, among others. Users can deploy and utilize these models seamlessly on the platform.
• Chip Heterogeneity: The platform is optimized for multiple chip architectures, ensuring efficient performance across different hardware setups. This includes support for NVIDIA, AMD, and other domestic chips like Iluvatar, Metax, MooreThreads, Biren, and Enflame.
• Unified Deployment: Infini-AI provides a unified interface for deploying and managing AI models, making it easier for developers to integrate AI capabilities into their applications.
• High-Performance Inference: The platform offers high-performance inference capabilities, enabling real-time responses for AI applications.
• Extensive Token Quota: Infini-AI provides a generous token quota for users, facilitating extensive testing and usage of AI models without immediate cost concerns.

How to Use Infini-AI

Getting started with Infini-AI is straightforward:

1. Create an Account: Visit the Infini-AI website and register for an account using your mobile number. No real-name authentication is required during registration.
2. Access the Console: After logging in, you can access the AI Studio console, where you can manage your AI models and resources.
3. Deploy a Development Machine: In the console, navigate to the “Development Machine” section and create a GPU-enabled development machine. Currently, only single-GPU machines are available for selection.
4. Access via SSH: Once your development machine is set up, you can access it via SSH using the provided credentials. This allows you to install necessary dependencies and run AI models locally.
5. Deploy Inference Services: You can deploy inference services using Docker containers. Infini-AI provides pre-configured Docker images for various AI models, which you can pull and run on your development machine.
6. Make API Calls: After setting up the inference service, you can make API calls to interact with the deployed AI models. Infini-AI supports OpenAI-compatible API endpoints, allowing you to integrate AI capabilities into your applications easily.

Pricing

Infini-AI offers a generous free trial for new users, providing a substantial token quota to facilitate extensive testing and usage of AI models. The platform operates on a token-based billing system, where users are charged based on the number of tokens processed during inference. Detailed pricing information can be found on the official pricing page.

Frequently Asked Questions (FAQs)

• Can I deploy my own models on Infini-AI? Yes, Infini-AI allows users to upload and deploy their own models. You can upload models trained via fine-tuning or LoRA and deploy them as online services. Refer to the official documentation for more details on uploading and deploying custom models.
• What is the maximum token usage per day? During the public beta phase, Infini-AI provides a daily token quota of up to 3 billion tokens, allowing users to make extensive use of AI models without immediate cost concerns.
• What chip architectures are supported? Infini-AI supports various chip architectures, including NVIDIA, AMD, Iluvatar, Metax, MooreThreads, Biren, and Enflame, ensuring efficient performance across different hardware setups.
• How can I monitor my token usage? Infini-AI provides a dashboard where users can monitor their token usage and manage their quotas effectively.