ml_reports

qoa4ml.reports.ml_reports

Classes

MLReport

MLReport manages the reporting of machine learning metrics and inference data.

Parameters:

Name	Type	Description	Default
client_config	ClientInfo	Configuration settings related to the client.	required

Attributes:

Name	Type	Description
client_config	ClientInfo	A deep copy of the client configuration.
init_time	float	The initialization time of the report.
previous_report	list[GeneralMlInferenceReport]	A list of previously processed reports.
report	GeneralMlInferenceReport	The current state of the report.

Methods:

Name	Description
reset	Reset the report to an initial state.
combine_stage_report	Combine metrics from the current and previous stage reports.
process_previous_report	Process and incorporate a previous report.
observe_metric	Observe and record a metric.
observe_inference	Observe and record inference data.
observe_inference_metric	Observe and record an inference-specific metric.
generate_report	Generate the report and optionally reset the current report state.

Functions

__init__(client_config)

Initialize an instance of MLReport.

Parameters:

Name	Type	Description	Default
client_config	ClientInfo	Configuration settings related to the client.	required

combine_stage_report(current_stage_report, previous_stage_report)

Combine metrics from the current and previous stage reports.

Parameters:

Name	Type	Description	Default
current_stage_report	dict	The current stage report containing metrics.	required
previous_stage_report	dict	The previous stage report containing metrics.	required

Returns:

Type	Description
dict	Combined stage report containing metrics from both reports.

generate_report(reset=True, corr_id=None)

Generate the report and optionally reset the current report state.

Parameters:

Name	Type	Description	Default
reset	bool	Whether to reset the report state after generating the report, default is True.	True
corr_id	Optional[str]	Correlation ID for the report, default is None.	None

Returns:

Type	Description
BaseReport	The generated report.

Notes

• Adds metadata such as client configuration, timestamp, and runtime to the report.
• Deep copies the current state of the report before optionally resetting it.

observe_inference(inference_value)

Observe and record inference data.

Parameters:

Name	Type	Description	Default
inference_value	Any	The value of the inference to be recorded.	required

Notes

• Raises a warning if inference data already exists for the current instance.

observe_inference_metric(metric)

Observe and record an inference-specific metric.

Parameters:

Name	Type	Description	Default
metric	Metric	The inference-specific metric to be recorded.	required

observe_metric(report_type, stage, metric)

Observe and record a metric.

Parameters:

Name	Type	Description	Default
report_type	ReportTypeEnum	The type of report being generated.	required
stage	str	The stage of the process in which the metric is recorded.	required
metric	Metric	The metric to be recorded.	required

Raises:

Type	Description
ValueError	If the stage name is empty or the report type is not handled.

process_previous_report(previous_report_dict)

Process and incorporate a previous report.

Parameters:

Name	Type	Description	Default
previous_report_dict	dict	Dictionary representation of a previous report.	required

Notes

• Service quality, data quality, and ML inference reports are combined with the current report.

reset()

Reset the report to an initial state.

Notes

• This method initializes a new report model and clears the list of previous reports.