D-Robotics Official Website
D-Robotics Official Website
English

D-Robotics-LLM Document Introduction

This document is intended for all developers using D-Robotics-LLM. It provides comprehensive guidance throughout your development process. To help you fully understand the overall usage workflow, D-Robotics recommends reading this document in the following order.

Below is a brief overview of each section. You may also navigate directly to specific chapters based on your needs.

1. LLM Document Introduction

This section provides you with brief introductions to and navigation links for relevant sections in the overall document, as well as the recommended reading order of the document.

  1. Product Introduction
SectionSection Introduction
OpenExplorer IntroductionThis section introduces D-Robotics-LLM to you, and also provides a brief overview of the release contents included in the package.
Development ProcessThis section provides a brief overview of the overall development process for D-Robotics-LLM.
  1. Environmental Deployment

This section outlines the pre-required environment deployment steps and details, you need to complete for the development environment.

  1. Usage Instructions
SectionSection Introduction
S100/S100P PlatformThis chapter provides a usage guide for the DeepSeek-R1-Distill-Qwen, InternLM2, Qwen2.5, and Qwen2.5-Omni models. It covers both basic and advanced development paths to help you gain a comprehensive understanding of the entire usage process.
  1. Tools Introduction
SectionSection Introduction
The oellm_build ToolThe oellm_build tool is provided by D-Robotics to map floating-point model to quantized model. It completes model quantization and compilation based on the original floating-point model, a json configuration file(optional), and calibration data(optional), and finally generates a deployable *.hbm model.
  1. API Introduction

This section introduces the API interfaces to help you better understand and use the API.

SectionSection Introduction
API IntroductionThis section uses a diagram of the basic API calling workflow to explain how to use API interfaces for LLM model loading, inference, and result processing.
Detailed API DescriptionThis section provides a detailed introduction to the API interfaces for LLM model inference, including content such as basic type definition, structure and enumeration types, callback function type, and main interface functions.
  1. Benchmark of Model Performance
SectionSection Introduction
Benchmark of DeepSeek-R1-Distill-Qwen Model PerformanceThis section presents the various DeepSeek-R1-Distill-Qwen model benchmark data obtained under specific test conditions, allowing you to clearly understand the model's performance.
Benchmark of InternLM2 Model PerformanceThis section presents the various InternLM2 model benchmark data obtained under specific test conditions, allowing you to clearly understand the model's performance.
Benchmark of Qwen2.5-Omni Model PerformanceThis section presents the various Qwen2.5-Omni model benchmark data obtained under specific test conditions, allowing you to clearly understand the model's performance.
Benchmark of Qwen2.5 Model PerformanceThis section presents the various Qwen2.5 model benchmark data obtained under specific test conditions, allowing you to clearly understand the model's performance.
  1. D-Robotics-LLM Algorithm Toolchain License Agreement

This section provides the D-Robotics-LLM algorithm toolchain license agreement. You must read it carefully before using the D-Robotics-LLM toolchain.