Overview
- Jupyter notebook-based development environment
- Single development environment—explore, train, deploy ML models via code
- Unified platform
- Seamless, scalable, sustainable, speedy—models have 80% fewer lines of code than competitors
Notebooks
- Google or user managed options
- User managed:
- Deep Learning VM Images
- Customizable environment
- Packaged with JupyterLab
- Pre-installed suite of packages e.g. TensorFlow, PyTorch
- GPU support
- Sync with GitHub
- Can run notebooks on Dataproc