AI & Deep learning docker container(s) to go...
Tensorflow: 2.0.0-alpha0
Fastai: 1.0.51
- Make sure the host system for running the container has CUDA 10.1 installed.
-
Install CUDA 10.1 on the host system https://developer.nvidia.com/cuda-toolkit
-
Install Docker: (docker-ce or docker-ee) https://docs.docker.com/install/
-
Install Nvidia-docker: https://github.com/NVIDIA/nvidia-docker
-
Checkout code and run container
git clone https://github.com/fastai/course-v3.git
cd course-v3
docker run --runtime=nvidia -it -p 8888:8888 -v $HOME/.fastai:/root/.fastai -v $(pwd):/code/fastai --ipc=host zerotosingularity/fastai_v3:latest- Kaggle (Optional) If you want to use the Kaggle command line tools:
- place the kaggle.json file in the ~/.kaggle/ folder
- Go to Kaggle.com -> "My Account" -> "Create New API Token"
- add another volume when starting the container: -v ~/.kaggle:/YOUR_USER/.kaggle
- YOUR_USER should be changed to your actual username, because the target volume needs to be an absolute path
- full example: docker run --runtime=nvidia -it -p 8888:8888 -v
$HOME/.fastai:/root/.fastai -v $ (pwd):/code/fastai -v ~/.kaggle:/YOUR_USER/.kaggle --ipc=host zerotosingularity/fastai_v3:latest
It uses three volumes:
- /root/.fastai: store the data for later use, so you don't have to redownload every time
- /root/.torch: store the data for later use, so you don't have to redownload every time
- /code/fastai: maps to the current (course-v3) repository, which lets you save changes over time, and simply pull updates
- (Optional Kaggle volume as described in 4. Kaggle (Optional))
This project is licensed under the MIT License - see the LICENSE.md file for details
- Inspired by: https://github.com/MattKleinsmith/dockerfiles/blob/master/fastai/Dockerfile;
- Learned from: Run Jupyter script was seen at: Floydhub - dl-docker: https://github.com/floydhub/dl-docker/blob/master/run_jupyter.sh;
- Thanks to: Seppe De Loore for the initial fastai installation steps when we needed them.