update readme
Browse files- README.md +1 -1
- README_en.md +1 -1
README.md
CHANGED
|
@@ -29,7 +29,7 @@ tasks:
|
|
| 29 |
|
| 30 |
## 介绍
|
| 31 |
|
| 32 |
-
**如意大模型（AI-Flow-Ruyi）**
|
| 33 |
|
| 34 |
|
| 35 |
|
|
|
|
| 29 |
|
| 30 |
## 介绍
|
| 31 |
|
| 32 |
+
**如意大模型（AI-Flow-Ruyi）** 是中国电信人工智能研究院 (TeleAI) 智传网（AI Flow）团队研发，是面向下一代“端-边-云”模型服务架构的**同源家族模型（Familial Model）** 。其核心在于大小模型共享同源参数，模型能基于早退出机制，根据问题复杂度调用不同参数规模的分支模型进行响应。各分支既可独立运行，又能依托同源特性实现信息共享与无缝切换，结合端-边-云分布式部署，完成家族大小模型协同，实现模型分布式推理效率大幅提升。
|
| 33 |
|
| 34 |
|
| 35 |
|
README_en.md
CHANGED
|
@@ -25,7 +25,7 @@ tasks:
|
|
| 25 |
|
| 26 |
## Introduction
|
| 27 |
|
| 28 |
-
**AI-Flow-Ruyi** is a **Familial Model** developed by the AI Flow team
|
| 29 |
|
| 30 |
|
| 31 |
|
|
|
|
| 25 |
|
| 26 |
## Introduction
|
| 27 |
|
| 28 |
+
**AI-Flow-Ruyi** is a **Familial Model** developed by the AI Flow team of the Institute of Artificial Intelligence (TeleAI), China Telecom. Designed for next-generation "Device-Edge-Cloud" model service architectures, its core innovation lies in **shared familial parameters** across large and small models. Leveraging an **early-exit mechanism**, the system dynamically routes queries to branch models of appropriate parameter sizes based on problem complexity. These branches operate independently while enabling **information sharing** and **seamless transitions** through their shared features. Combined with distributed Device-Edge-Cloud deployment, this facilitates **collaborative inference** within the model family, significantly enhancing distributed reasoning efficiency.
|
| 29 |
|
| 30 |
|
| 31 |
|