The docker image has been released to DockerHub
NVIDIA DRIVER VERSION >= 418.48 (compatible with CUDA10.0 according to here)
nvidia-docker
ensure echo $USER output your username first. If not, set it with export USER="your_user_name"
bash start_docker.sh # start a container named $USER_deep_docker
bash enter_docker.sh # enter the docker
nvidia/cuda:10.0-cudnn7-devel-ubuntu18.04
| Library | version | Note |
|---|---|---|
| Tensorflow | 2.2.0 | GPU-version, with Tensorboard |
| pytorch | 1.5.0 | with TensorboardX, torchvision |
| apex | 0.1 | |
| nvidia-dali | 0.21.0 |
numpy scipy scikit-image Pillow opencv-3.3.1 matplotlib ipython jupyter pandas
cmake automake oh-my-zsh(install during starting a container)