fetch_metrics()

Python package: neptune-fetcher

Returns raw values for the requested metrics. The returned values don't include any aggregation, approximation, or interpolation.

Fetch experiment metrics

You can filter the results by:

• Experiments: Specify which experiments to search.
• Attributes: Only list attributes that match certain criteria.

Parameters

Name	Type	Default	Description
experiments	Union[str, Filter]	-	A filter specifying which experiments to include: a regex that the experiment name must match, or a Filter object.
attributes	Union[str, AttributeFilter]	-	A filter specifying which attributes to include in the table: a regex that the attribute name must match, or an AttributeFilter object. If AttributeFilter.aggregations is set, an exception will be raised as they're not supported in this function.
include_time	Literal["absolute"], optional	None	Whether to include absolute timestamp. If set, each metric column has an additional sub-column with requested timestamp values.
step_range	Tuple[float, float]	None	A tuple specifying the range of steps to include. Can represent an open interval.
lineage_to_the_root	bool	True	If True, includes all points from the complete experiment history. If False, only includes points from the selected experiment.
tail_limit	int, optional	None	From the tail end of each series, how many points to include at most.
type_suffix_in_column_names	bool	False	If True, columns of the returned DataFrame will be suffixed with ":<type>". For example: "attribute1:float_series", "attribute1:string". If set to False, the method throws an exception if there are multiple types under one path.
include_point_previews	bool, optional	False	If set to True, metric previews are included in the fetched data frame.
context	Context, optional	None	Which project and API token to use for the fetching operation. Useful for switching projects.

Example

Fetch loss metrics from experiments matching a regex, including point previews and only values logged from step 1000 onward:

import neptune_fetcher.alpha as npt


npt.fetch_metrics(
    experiments=r"seagull.*_estimated$",
    attributes=r"^loss/.*",
    step_range=(1000.0, None),
    include_point_previews=True,
)

Output

                            path     loss/train
                                     is_preview     preview_completion	 value
               experiment   step			
seagull-45xc099_estimated   1000.0        False	                   1.0	 0.193153
                            1001.0        False	                   1.0	 0.166237
                            1002.0        False	                   1.0	 0.12602
...

Fetch run metrics

You can filter the results by:

• Runs: Specify which runs to search.
• Attributes: Only list attributes that match certain criteria.

Parameters

Name	Type	Default	Description
runs	Union[str, Filter]	-	A filter specifying which runs to include: a regex that the run ID must match, or a Filter object.
attributes	Union[str, AttributeFilter]	-	A filter specifying which attributes to include in the table: a regex that the attribute name must match, or an AttributeFilter object. If AttributeFilter.aggregations is set, an exception will be raised as they're not supported in this function.
include_time	Literal["absolute"], optional	None	Whether to include absolute timestamp. If set, each metric column has an additional sub-column with requested timestamp values.
step_range	Tuple[float, float]	None	A tuple specifying the range of steps to include. Can represent an open interval.
lineage_to_the_root	bool	True	If True, includes all points from the complete run history. If False, only includes points from the selected runs.
tail_limit	int, optional	None	From the tail end of each series, how many points to include at most.
type_suffix_in_column_names	bool	False	If True, columns of the returned DataFrame will be suffixed with ":<type>". For example: "attribute1:float_series", "attribute1:string". If set to False, the method throws an exception if there are multiple types under one path.
include_point_previews	bool, optional	False	If set to True, metric previews are included in the fetched data frame.
context	Context, optional	None	Which project and API token to use for the fetching operation. Useful for switching projects.

Example

Fetch accuracy metrics from runs matching a regex, including the last 3 values from each series:

from neptune_fetcher.alpha import runs


runs.fetch_metrics(
    runs=r"marigold",
    attributes=r"accuracy",
    tail_limit=3,
)

Output

                                    accuracy
                              run   step
arrogant-millipede+marigold-finch   49.0	0.830062
                                    50.0	0.828428
                                    51.0	0.825925
 marigold-finch+realistic-dolphin   31.0	0.970358
                                    32.0	0.986717
                                    33.0	0.971492
 marigold-finch+thundering-mantis   29.0	0.925642
                                    30.0	0.963742
                                    31.0	0.970358