README.rst

Metric Reporting Python Package for CloudML Hypertune

Helper Functions for CloudML Engine Hypertune Services.

Prerequisites

• Google CloudML Engine Overview.
• Google CloudML Engine Hyperparameter Tuning Overview.

Installation

Install via pip:

pip install cloudml-hypertune

Usage

import hypertune

hpt = hypertune.HyperTune()
hpt.report_hyperparameter_tuning_metric(
    hyperparameter_metric_tag='my_metric_tag',
    metric_value=0.987,
    global_step=1000)

By default, the metric entries will be stored to /tmp/hypertune/output.metrics in json format:

{"global_step": "1000", "my_metric_tag": "0.987", "timestamp": 1525851440.123456, "trial": "0"}

Licensing

• Apache 2.0