《峡谷小狐仙-多模态角色扮演游戏助手》复现流程

YongXie66/Honor-of-Kings_RolePlay: The Role Playing Project of Honor-of-Kings Based on LnternLM2。峡谷小狐仙--王者荣耀领域的角色扮演聊天机器人，结合多模态技术将英雄妲己的形象带入大模型中。 (github.com)
 

https://github.com/chg0901/Honor_of_Kings_Multi-modal_Dataset

https://github.com/YongXie66/Honor-of-Kings_RolePlay

[InternLM][书生浦语][WAIC]峡谷小狐仙--多模态角色扮演游戏助手[松龄后裔 团队] - 知乎: https://zhuanlan.zhihu.com/p/708285503
 

 1、创建环境

git clone https://github.com/YongXie66/Honor-of-Kings_RolePlay.git

conda create -n hok-roleplay python=3.10

conda activate hok-roleplay

pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118


cd Honor-of-Kings_RolePlay

pip install -r requirements.txt

conda install -q ffmpeg

2、下载模型（下载时间太长，耐心等等)

apt install git
apt install git-lfs

# LLM, funasr, gpt_sovits, sadtalker 相关模型下载
git clone https://code.openxlab.org.cn/YongXie66/DaJi_RolePlay.git ./DaJi_RolePlay

cd DaJi_RolePlay/InternLM2/InternLM2_7b/
bash download.sh

mv ./DaJi_RolePlay/InternLM2/InternLM2_7b/InternLM2/InternLM2_7b/* /Honor-of-Kings_RolePlay/InternLM2/InternLM2_7b

bash FunASR/download.sh

mv ./DaJi_RolePlay/checkpoints/* ./checkpoints/
mv ./DaJi_RolePlay/gfpgan/* ./gfpgan/

mv ./DaJi_RolePlay/FunASR/* ./Honor-of-Kings_RolePlay/FunASR/
mv ./DaJi_RolePlay/GPT_SoVITS/pretrained_models/*  ./Honor-of-Kings_RolePlay/GPT_SoVITS/pretrained_models

 

 

# 生成 RAG 依赖的 Chroma 数据库
python Honor-of-Kings_RolePlay/rag/generate_chroma_db.py

3、Web UI 启动 !

python webui.py

4、总结

1、移动的文件太多太大，要检测清楚所有文件夹下文件齐全。

2、配置环境库按照 requirements.txt 安装

gradio== 4.29.0

下面这个细节没改！！！