Skip to content
/ vizier Public

Python-based research interface for blackbox and hyperparameter optimization, based on the internal Google Vizier Service.

License

Notifications You must be signed in to change notification settings

google/vizier

Folders and files

NameName
Last commit message
Last commit date

Latest commit

vizier-teamcopybara-github
vizier-team
and
Feb 14, 2025
4b94703 · Feb 14, 2025
Feb 1, 2025
Jan 6, 2025
Jan 7, 2025
Feb 14, 2025
Sep 7, 2023
Oct 6, 2022
Mar 1, 2022
Nov 11, 2024
Jan 8, 2024
Jan 23, 2025
Jan 23, 2025
Oct 21, 2024
Jan 13, 2023
Jan 27, 2023
Jan 31, 2025
Jan 12, 2024
Apr 12, 2024

Open Source Vizier: Reliable and Flexible Black-Box Optimization.

PyPI version Continuous Integration Docs

Google AI Blog | Getting Started | Documentation | Installation | Citing and Highlights

What is Open Source (OSS) Vizier?

OSS Vizier is a Python-based service for black-box optimization and research, based on Google Vizier, one of the first hyperparameter tuning services designed to work at scale.


OSS Vizier's distributed client-server system. Animation by Tom Small.

Getting Started

As a basic example for users, below shows how to tune a simple objective using all flat search space types:

from vizier.service import clients
from vizier.service import pyvizier as vz

# Objective function to maximize.
def evaluate(w: float, x: int, y: float, z: str) -> float:
  return w**2 - y**2 + x * ord(z)

# Algorithm, search space, and metrics.
study_config = vz.StudyConfig(algorithm='DEFAULT')
study_config.search_space.root.add_float_param('w', 0.0, 5.0)
study_config.search_space.root.add_int_param('x', -2, 2)
study_config.search_space.root.add_discrete_param('y', [0.3, 7.2])
study_config.search_space.root.add_categorical_param('z', ['a', 'g', 'k'])
study_config.metric_information.append(vz.MetricInformation('metric_name', goal=vz.ObjectiveMetricGoal.MAXIMIZE))

# Setup client and begin optimization. Vizier Service will be implicitly created.
study = clients.Study.from_study_config(study_config, owner='my_name', study_id='example')
for i in range(10):
  suggestions = study.suggest(count=2)
  for suggestion in suggestions:
    params = suggestion.parameters
    objective = evaluate(params['w'], params['x'], params['y'], params['z'])
    suggestion.complete(vz.Measurement({'metric_name': objective}))

Documentation

OSS Vizier's interface consists of three main APIs:

  • User API: Allows a user to optimize their blackbox objective and optionally setup a server for distributed multi-client settings.
  • Developer API: Defines abstractions and utilities for implementing new optimization algorithms for research and to be hosted in the service.
  • Benchmarking API: A wide collection of objective functions and methods to benchmark and compare algorithms.

Additionally, it contains advanced API for:

  • Tensorflow Probability: For writing Bayesian Optimization algorithms using Tensorflow Probability and Flax.
  • PyGlove: For large-scale evolutionary experimentation and program search using OSS Vizier as a distributed backend.

Please see OSS Vizier's ReadTheDocs documentation for detailed information.

Installation

Quick start: For tuning objectives using our state-of-the-art JAX-based Bayesian Optimizer, run:

pip install google-vizier[jax]

Advanced Installation

Minimal installation: To install only the core service and client APIs from requirements.txt, run:

pip install google-vizier

Full installation: To support all algorithms and benchmarks, run:

pip install google-vizier[all]

Specific installation: If you only need a specific part "X" of OSS Vizier, run:

pip install google-vizier[X]

which installs add-ons from requirements-X.txt. Possible options:

  • requirements-jax.txt: Jax libraries shared by both algorithms and benchmarks.
  • requirements-tf.txt: Tensorflow libraries used by benchmarks.
  • requirements-algorithms.txt: Additional repositories (e.g. EvoJAX) for algorithms.
  • requirements-benchmarks.txt: Additional repositories (e.g. NASBENCH-201) for benchmarks.
  • requirements-test.txt: Libraries needed for testing code.

Developer installation: To install up to the latest commit, run:

pip install google-vizier-dev[X]

Check if all unit tests work by running run_tests.sh after a full installation. OSS Vizier requires Python 3.10+, while client-only packages require Python 3.8+.

Citing and Highlights

Citing Vizier: Please consider citing the appropriate paper(s): Algorithm, OSS Package, and Google System if you found any of them useful.

Highlights: We track notable users and media attention - let us know if OSS Vizier was helpful for your work.

Thanks!

@article{gaussian_process_bandit,
  author       = {Xingyou Song and
                  Qiuyi Zhang and
                  Chansoo Lee and
                  Emily Fertig and
                  Tzu-Kuo Huang and
                  Lior Belenki and
                  Greg Kochanski and
                  Setareh Ariafar and
                  Srinivas Vasudevan and
                  Sagi Perel and
                  Daniel Golovin},
  title        = {The Vizier Gaussian Process Bandit Algorithm},
  journal      = {Google DeepMind Technical Report},
  year         = {2024},
  eprinttype    = {arXiv},
  eprint       = {2408.11527},
}

@inproceedings{oss_vizier,
  author    = {Xingyou Song and
               Sagi Perel and
               Chansoo Lee and
               Greg Kochanski and
               Daniel Golovin},
  title     = {Open Source Vizier: Distributed Infrastructure and API for Reliable and Flexible Black-box Optimization},
  booktitle = {Automated Machine Learning Conference, Systems Track (AutoML-Conf Systems)},
  year      = {2022},
}

@inproceedings{google_vizier,
  author    = {Daniel Golovin and
               Benjamin Solnik and
               Subhodeep Moitra and
               Greg Kochanski and
               John Karro and
               D. Sculley},
  title     = {Google Vizier: {A} Service for Black-Box Optimization},
  booktitle = {Proceedings of the 23rd {ACM} {SIGKDD} International Conference on
               Knowledge Discovery and Data Mining, Halifax, NS, Canada, August 13
               - 17, 2017},
  pages     = {1487--1495},
  publisher = {{ACM}},
  year      = {2017},
  url       = {https://doi.org/10.1145/3097983.3098043},
  doi       = {10.1145/3097983.3098043},
}