AutoML Examples

Compilation of programs for various topics in AutoML using the Ray framework from UC Berkeley's RiseLab

Examples:

• set_env_vars_example.py
  -- how to start off actors in Ray Tune with different environment variables
• pytorch_hparams.py
  -- classic example of finding optimal hyperparameters for simple mnist classifying CNN
• naive_pytorch_nas.py
  -- very simple example of naive neural architecture search with simple mnist classifying CNN
  -- naive because this is effectively just another hyperparameter sweep, not actual architecture search
• tune_resnet50.py
  -- tuning inference time for resnet50 with imagenet weights.
  -- uses Intel Optimized tensorflow

References

• Ray documentation
• Intro to Neural Architecture Search