MindIE

  • Large language model (LLM) inference scenario:
    • Using the open-source inference engine vLLM: Ascend supports open-source inference engines, such as vLLM, to accelerate inference.
    • MindIE text generation: MindIE LLM enables text generation inference and accelerates the E2E serving inference process.
  • Multimodal inference scenario:
    • MindIE vision generation: MindIE SD enables high-performance vision generation inference.
  • Release Notes

    Describes the version mapping, new features, modified features, deleted features, known issues, resolved issues, and fixed vulnerabilities in the current MindIE version.

  • Quick Start

    Describes the overall architecture of MindIE and how to get started with model inference journeys.

  • List of the Supported Models

    Describes the models supported by MindIE.

Using the vLLM Open-Source Inference Engine

MindIE Text Generation

  • Getting started with MindIE text generation inference

    Describes how to use MindIE LLM on the Atlas 800I A2 inference server to quickly perform text generation inference.

  • List of the Supported Models

    Contains LLMs supported by MindIE.

  • MindIE Installation Guide

    Describes how to install, upgrade, and uninstall MindIE.

  • MindIE LLM Development Guide

    Describes the model inference process, features such as quantization, long sequence, and multi-machine functions, as well as application programming interface (API) usage. MindIE LLM is a large language model inference component of MindIE.

  • MindIE Motor Development Guide

    Describes the inference serving capabilities of MindIE Motor, functions of each component, and application scenarios. MindIE Motor is an inference serving framework designed for general-purpose model scenarios, including MindIE Service Tools, MindIE Client, and cluster management components (Deployer, Controller, and Coordinator).

MindIE Vision Generation

References