devtography/swift-tf repository overview

⁠Swift for TensorFlow - Docker

⁠An Ubuntu 18.04 Docker image for Swift for TensorFlow⁠.

⁠You can find the Docker Hub repo here: https://hub.docker.com/r/devtography/swift-tf⁠

⁠Supported tags & respective Swift for TensorFlow version

* Append -cuda postfix for GPU supported images

⁠Usage

⁠Pull the Docker image from Docker Hub:

# CPU only
docker pull devtography/swift-tf

# With CUDA support
docker pull devtography/swift-tf:latest-cuda

⁠Create a container from the image and run it:

docker run -it devtography/swift-tf

If you want to run the Swift REPL you will need to run the container with additional privileges:

docker run --security-opt seccomp=unconfined -it devtography/swift-tf

For GPU support on Linux, install NVIDIA Docker support⁠ and run a GPU-enabled Swift for TensorFlow image:

• Take note of your Docker version with docker -v. Versions earlier than 19.03 require nvidia-docker2 and the --runtime=nvidia flag. On versions including and after 19.03, you will use the nvidia-container-toolkit package and the --gpus all flag.

# Docker version >= 19.03
docker run --gpus all -it devtography/swift-tf:latest-cuda

# Docker version < 19.03
docker run --runtime=nvidia -it devtography/swift-tf:latest-cuda

Note from official Swift for TensorFlow installation doc⁠:

⁠Remarks

⁠Jupyter Notebook support

Jupyter Notebook support for Swift is NOT included in the docker images. If you would like to try Swift for TensorFlow⁠ in a Jupyter Notebook, Google's Colaboratory⁠ provides a free an no setup required environment that runs entirely in the cloud with GPU support.

For more details, please follow the usage⁠ document of Swift for TensorFlow⁠.

⁠Contributions

Contributions via pull requests are welcome and encouraged :)

⁠License

swift-tf-docker is licensed under the Apache License, Version 2.0⁠.