jurgisp/expa

Metric logging and analysis tool for ML experiments

eXpa

Metric logging and analysis tool for ML experiments.

The components are:

• Scalable data store backed by PostgreSQL
• REST API for logging and retrieving data
• Python client library expa for sending metrics
• Powerful UI for metric visualization

This is a pre-release work in progress, expect breaking changes!

Getting started

Start development server locally

git clone https://github.com/jurgisp/expa.git && cd expa
PORT=8000 docker-compose up

Use expa client to push metrics

!pip install expa
import expa
logger = expa.Logger('experiment', 'run', api_url='http://localhost:8000/api')
logger.log_params({'param': 'value'})
for step in range(1, 101):
  logger.log({'loss': 1/step}, step)

Visualize metrics

Open localhost:8000

Security

The API does not have any user authentication built-in. Be sure to add an auth layer if you are planning to deploy the service on publicly accessible network.

Contributing

See CONTRIBUTING.md for details.

License

Apache 2.0; see LICENSE for details.

Disclaimer

This project is not an official Google project. It is not supported by Google and Google specifically disclaims all warranties as to its quality, merchantability, or fitness for a particular purpose.

This product is not intended or suitable for storing private data.