ms | modelscope conda方式安装

### conda虚拟环境构建
# anaconda安装过程略过
# 创建modelscope_py37 conda虚拟环境
conda create -n modelscope_py38 python=3.8.5

# 激活虚拟环境
conda activate modelscope_py38

### 安装pytorch和tensorflow
# 安装pytorch
pip install -i https://pypi.douban.com/simple torch torchvision torchaudio

# 安装tensorflow
pip install -i https://pypi.douban.com/simple tensorflow

### modelscope library安装
# 安装modelscope核心框架 ==> 使用modelscope模型和数据集的基础能力
pip install -i https://pypi.douban.com/simple modelscope

# 安装`多模态`依赖
pip install -i https://pypi.douban.com/simple "modelscope[multi-modal]" 

# 安装`nlp`依赖
pip install -i https://pypi.douban.com/simple "modelscope[nlp]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html

# 安装`cv`依赖
pip install -i https://pypi.douban.com/simple "modelscope[cv]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html
pip uninstall mmcv # 如果已经安装过简装版本的mmcv，请先卸载
pip install -i https://pypi.douban.com/simple -U openmim 
mim install mmcv-full

# 安装`语音`依赖
pip install -i https://pypi.douban.com/simple "modelscope[audio]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html

# 安装`科学计算`依赖
pip install -i https://pypi.douban.com/simple "modelscope[science]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html