Skip to content

Latest commit

 

History

History
58 lines (39 loc) · 2.38 KB

contributing.md

File metadata and controls

58 lines (39 loc) · 2.38 KB
Raw

Contributing

Get started with Kepler Model Server.

PR Hands-on

  • Create related issue with your name assigned first (if not exist).

  • Set required secret and environment for local repository test if needed. Check below table.

Objective Required Secret Required Environment
Push to private repo BOT_NAME, BOT_TOKEN IMAGE_REPO
Change on base image BOT_NAME, BOT_TOKEN IMAGE_REPO
Save data/models to AWS COS AWS_ACCESS_KEY_ID,AWS_SECRET_ACCESS_KEY,AWS_REGION

Improve components in training pipelines

Learn more details about Training Pipeline

Introduce new feature group

  • Define new feature group name FeatureGroup and update metric list map FeatureGroups in train types

Introduce new energy sources

  • Define new energy source map PowerSourceMap in train types

Improve preprocessing method

  • extractor: convert from numerically aggregated metrics to per-second value
  • isolator: isolate background (idle) power from the collected power

Introduce new learning method

  • trainer: apply learning method to build a model using extracted data and isolated data

Model training

Learn more details about model training

Introduce new benchmarks

The new benchmark must be supported by CPE operator for automation. Find examples.

CPE-based (deprecated)

Benchmark CR has a dependency on BenchmarkOperator. Default BechmarkOperator is to support batch/v1/Job API.

Tekton

Create workload Task and provide example Pipeline to run.

Add new trained models

TBD

Source improvement

Any improvement in src and cmd.

Test and CI improvement

Any improvement in tests, dockerfiles, manifests and .github/workflows

Documentation

Detailed documentation should be posted to kepler-doc repository.