Google Deep Learning Containers: DevOps for Data Science
Google Cloud Platform recently announced the beta release of Google Deep Learning Containers, a family of preconfigured Docker containers outfitted with deep learning frameworks and tools. The freely-available containers (with a Google Cloud Platform account) target a common pain point for data scientists and developers -- the complexity involved in setting up and configuring deep learning projects.
Deep Learning Containers support popular ML frameworks like PyTorch, TensorFlow 2.0, and TensorFlow 1.13, and come equipped with preconfigured Jupyter and Google Kubernetes Engine (GKE) clusters. The tooling, Google says, provides a consistent environment for testing and deploying applications. Deep Learning Containers can be run locally with Docker, Docker Compose or Kubernetes.
"Each container provides a Python3 environment consistent with the corresponding Deep Learning VM, including the selected data science framework, conda, the NVIDIA stack for GPU images (CUDA, cuDNN, NCCL), and a host of other supporting packages and tools," wrote Google Software Engineer Mike Cheng in a blog post announcing the launch.
The Google release comes about three months after Amazon launched AWS Deep Learning Containers, another library of pre-built Docker images supporting widely deployed deep learning frameworks. The releases reflect an ongoing effort in the sector to broaden the appeal of ML beyond traditional, data science organizations.
"Productionizing your workflow requires not only developing the code or artifacts you want to deploy, but also maintaining a consistent execution environment to guarantee reproducibility and correctness," Cheng wrote. "If your development strategy involves a combination of local prototyping and multiple cloud tools, it can often be frustrating to ensure that all the necessary dependencies are packaged correctly and available to every runtime."
Michael Desmond is an editor and writer for 1105 Media's Enterprise Computing Group.