我参加NVIDIA Sky Hackathon

环境安装

基础环境：

显卡驱动：

sudo apt-get install nvidia-driver-515

reboot

切换独显：

sudo prime-select nvidia reboot

miniconda(轻量化anaconda)：

#用国内清华源

export DL_SITE=https://mirrors.tuna.tsinghua.edu.cn/anaconda/miniconda

# 下载安装包

wget -c $DL_SITE/Miniconda3-py38_4.10.3-Linux-x86_64.sh

# 执行安装，全部按照预设路径与”yes”选项

bash Miniconda3-py38_4.10.3-Linux-x86_64.sh

# 启动

Conda source ~/.bashrc

安装 Jupyter Lab

pip install jupyter jupyterlab

# 设置登录密码

export PW='自行提供'

python3 -c "from notebook.auth.security import set_password;set_password('$PW','$HOME/.jupyter/jupyter_notebook_config.json')"

#安装CUDA(依据自身显卡版本安装 推荐30系显卡安装11.2)

nvidia-smi

#查看显卡驱动安装是否成功 CUDA显示为支持的最高版本可向下兼容

 

ubuntu18.04安装显卡驱动+CUDA+CUDNN_zhA____的博客-CSDN博客

 #安装教程

安装pytorch：

https://blog.csdn.net/gg864461719/article/details/112259806?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522165376105516782248598679%2522% 252C%2522scm%2522%253A%252220140713.130102334.pc%255Fall.%2522%257D&request_id=165376105516782248598679&biz_id=0

&utm_medium=distribute.pc_search_result.none-task-blog-2~all~first_rank_ecpm_v1~hot_rank-7-112259806-null-null.142^v11^control,157^v12

^control&utm_term=pytorch%E5%AE%89%E8%A3%85&spm=1018.2226.3001.4187

nemo环境安装

pip换源

pip config set global.index-url Simple Index

nemo安装

sudo apt-get update && sudo apt-get install -y libsndfile1 ffmpeg pip install Cython

pip install --user pytest-runner pip install rosa numpy==1.19.4 pip install torchmetrics==0.6.0

pip install nemo_toolkit[all]==1.4.0 pip install ASR-metrics

检测 Nemo

python

>>>import nemo

>>>import nemo.collections.asr as nemo_asr

>>>import nemo.collections.tts as nemo_tts

若没有报错表示安装成功（waring 不用理会）