[HpOpt] Hyper-parameter Optimization

Python library of automatic hyper-parameter optimization

Design principles

• Easy to use configuration
  • Support for automatic configuration
• Distributed execution on multi-node
  • Communication via persistent storage
    • Support for pause/resume
  • Decide to go or stop in the report method
    • Making a decision without a central controller
• General-purpose design
  • Support for sequential/parallel execution

Quick Start

• Installation

  $ pip install -r requirements.txt
  $ pip install -v -e .

• Sample codes

  • List of sample codes

    • Sequential Execution

      • toy_test.py
      • pytorch_test.py

    • Parallel Execution

      • toy_test_parallel.py
      • pytorch_parallel_test.py

  • How to run sample codes

    • Basic code to find an optimal point of skopt.benchmarks.branin

      $ cd samples
      $ pip install -r requirements.txt
      $ python toy_test.py