项目概述
The official code implementation for “Cache-to-Cache: Direct Semantic Communication Between Large Language Models”
项目地址
https://github.com/thu-nics/C2C
项目页面预览

关键指标
- Stars:314
- 主要语言:Python
- License:Apache License 2.0
- 最近更新:2026-01-14T01:42:55Z
- 默认分支:main
本站高速下载(国内可用)
- 源码压缩包下载:点击下载(本站镜像)
- SHA256:a2dedda4d2c6834c9ce8dad7c815e062bd072ad1df6691752573a215ec717da1
安装部署要点(README 精选)
Environment Setup
Create a new environment:
conda create -n rosetta python=3.10
conda activate rosetta
Install the package:
pip install -e .
For training and evaluation, install additional dependencies:
pip install -e ".[training,evaluation]"
Run chat example
We provide an interactive chat example to demonstrate cache-to-cache communication with pre-trained projectors in script/playground/live_chat_example.py.
# Single sharer
python script/playground/live_chat_example.py --checkpoint_dir path/to/checkpoint
# Multiple sharers (teacher models read from each checkpoint's config.json)
python script/playground/live_chat_example.py --checkpoint_dir path/to/ckpt1 path/to/ckpt2
常用命令(从 README 提取)
conda create -n rosetta python=3.10
conda activate rosetta
pip install -e .
pip install -e ".[training,evaluation]"
通用部署说明(适用于大多数项目)
- 下载源码并阅读 README
- 安装依赖(pip/npm/yarn 等)
- 配置环境变量(API Key、模型路径、数据库等)
- 启动服务并测试访问
- 上线建议:Nginx 反代 + HTTPS + 进程守护(systemd / pm2)
免责声明与版权说明
本文仅做开源项目整理与教程索引,源码版权归原作者所有,请遵循对应 License 合规使用。
© 版权声明
文章版权归作者所有,未经允许请勿转载。
THE END








暂无评论内容