Runs

Description

A lazy iterator of Run objects associated with a project and optional filter.

Runs are retrieved in pages from the W&B server as needed.

This is generally used indirectly using the Api.runs namespace.

Args

• client: Legacy GraphQL client retained for API compatibility.
• entity: (str) The entity (username or team) that owns the project.
• project: (str) The name of the project to fetch runs from.
• filters: (Optional[Dict[str, Any]]) A dictionary of filters to apply to the runs query.
• order: (str) Order can be created_at, heartbeat_at, config.*.value, or summary_metrics.*. If you prepend order with a + order is ascending (default). If you prepend order with a - order is descending. The default order is run.created_at from oldest to newest.
• per_page: (int) The number of runs to fetch per request (default is 50).
• include_sweeps: (bool) Whether to include sweep information in the runs. Defaults to True.
• lazy: (bool) Whether to lazily load run data or fetch full data immediately.
• service_api: (Optional[ServiceApi]) An optional ServiceApi instance for making additional API calls, such as fetching run history exports.

Methods

method histories

self,
samples: 'int' = 500,
keys: 'list[str] | None' = None,
x_axis: 'str' = '_step',
format: "Literal['default', 'pandas', 'polars']" = 'default',
stream: "Literal['default', 'system']" = 'default'

Return sampled history metrics for all runs that fit the filters conditions.

Arguments

• samples: The number of samples to return per run
• keys: Only return metrics for specific keys
• x_axis: Use this metric as the xAxis defaults to _step
• format: Format to return data in, options are “default”, “pandas”, “polars”
• stream: “default” for metrics, “system” for machine metrics

method next

self

Return the next item from the iterator. When exhausted, raise StopIteration

method update_variables

self

Update the query variables for the next page fetch.

method upgrade_to_full

self

Upgrade this Runs collection from lazy to full mode.

This switches to fetching full run data and upgrades any already-loaded Run objects to have full data. Uses parallel loading for better performance when upgrading multiple runs.