mlsync.engine package

Submodules

mlsync.engine.diff module

mlsync.engine.diff.diff(report_old, report_new)

Generate the diff report

MLSync reports can change the following ways: 1. Older experiments/runs were deleted: Supported 2. Newer experiments/runs were added: Supported 3. Newer experiments/runs have different metrics 4. Older experiments/runs have changed

MLSync reports can NOT change the following ways: 1. Older experiments/runs have different metrics

Parameters

  • report_old – the old report

  • report_new – the new report

mlsync.engine.sync module

class mlsync.engine.sync.Sync(report_format, producer, consumer, **kwargs)

Bases: object

Main class that runs the sync process.

Instantiates producer and destination APIs. Check docs for more details.

Parameters

  • report_format (str) – Path to the report format file in YAML format (see docs for more details)

  • producer (str) – Name of the producer API (e.g., mlflow)

  • consumer (str) – Name of the consumer API (e.g., notion)

Keyword Arguments

  • mlflow_uri (str) – MLFlow URI during the run (Optional)

  • notion_token (str) – Notion token (Optional)

  • notion_page_id (str) – Notion page ID (Optional)

Raises

  • NotImplementedError – If the producer or destination is not supported

  • ValueError – If required configurations are not provided (e.g., mlflow_uri, notion_token, notion_page_id)

sync(refresh_rate)

Sync between the producer and the destination.

Creates a diff report whenever there is a difference between the producer and the destination. Then the diff report is uploaded to the destination. We do not update the producer for any changes. The sync process runs in a loop until the user stops it. Refresh rate is an argument.

Parameters

refresh_rate (int) – Refresh rate in seconds

Module contents