安装¶
硬件环境¶
Minimum
4 GB system RAM
4 GB of GPU RAM
Single core CPU
1 GPU
50 GB of HDD space
*Recommended
32 GB system RAM
32 GB of GPU RAM
8 core CPU
4 GPUs
100 GB of SSD space
软件环境¶
Ubuntu 18.04 LTS
NVIDIA GPU Cloud account and API key
- https://ngc.nvidia.com/- docker-ce
- nvidia-docker2
NVIDIA GPU driver v410.xx or above
注意:推荐使用DeepStream 5.0
安装¶
# 登录账户
$ docker login nvcr.io
Username: $oauthtoken -- 就输这个就好了
Password: API_KEY -- 替换成NGC的api-key
#下载容器
docker pull nvcr.io/nvidia/tlt-streamanalytics:<version>
最新版本容器参考:Transfer Learning Toolkit for Video Streaming
Jupyter Notebook¶
在容器内部,TLT
提供了jupyter notebook
文件,用于详细说明整体训练过程
# 启动容器,开放8888端口
$ docker run --runtime=nvidia -it -v /home/<username>/tlt-experiments:/workspace/tlt-experiments -p 8888:8888 nvcr.io/nvidia/tlt-streamanalytics:<version>
# 进入/workspace/examples文件夹,启动jupyter notebook
$ jupyter notebook --ip 0.0.0.0 --allow-root
Note:修改<username>
和<version>