Feature Monitoring Window Configuration#
MonitoringWindowConfig #
row_percentage property writable #
row_percentage: float | None
The percentage of rows to fetch and compute the statistics on. Only used for windows of type ROLLING_TIME and ALL_TIME.
specific_value property writable #
specific_value: float | None
The specific value to use as reference. Only used for windows of type SPECIFIC_VALUE.
time_offset property #
time_offset: str | None
The time offset from the current time to the start of the time window. Only used for windows of type ROLLING_TIME.
training_dataset_version property writable #
training_dataset_version: int | None
The version of the training dataset to use as reference. Only used for windows of type TRAINING_DATASET.
window_config_type property writable #
window_config_type: WindowConfigType
Type of the window. It can be one of ALL_TIME, ROLLING_TIME, TRAINING_DATASET or SPECIFIC_VALUE.
window_length property writable #
window_length: str | None
The length of the time window. Only used for windows of type ROLLING_TIME.
__init__ #
__init__(
id: int | None = None,
window_config_type: str
| WindowConfigType
| None = WindowConfigType.SPECIFIC_VALUE,
time_offset: str | None = None,
window_length: str | None = None,
training_dataset_version: int | None = None,
specific_value: float | None = None,
row_percentage: float | None = None,
**kwargs,
)
Configuration to define the slice of data to compute statistics on.
Example
# Detection or reference window
## Rolling Time Window
monitoring_window_config = MonitoringWindowConfig(
time_offset="1d", # data inserted up to 1 day ago
window_length="1h", # data inserted until an one hour after time_offset
row_percentage=0.2, # include only 20% of the rows when computing statistics
)
## Rolling Commit Window (not supported yet)
monitoring_window_config = MonitoringWindowConfig(
commit_offset=10, # data inserted up to 10 commits ago
commit_num=5, # include 5 commits after commit_offset
)
# Only available for reference window
## Specific value
monitoring_window_config = MonitoringWindowConfig(specific_value=0.5)
## Training dataset
monitoring_window_config = MonitoringWindowConfig(
training_dataset_version=my_training_dataset.version
)
| PARAMETER | DESCRIPTION |
|---|---|
id | int, optional The id of the monitoring window config. TYPE: |
window_config_type | str, optional The type of the monitoring window config. One of ROLLING_TIME, TRAINING_DATASET, SPECIFIC_VALUE. TYPE: |
time_offset | str, optional The time offset of the monitoring window config. Only used for INSERT and SNAPSHOT window config types. TYPE: |
window_length | str, optional The window length of the monitoring window config. Only used for INSERT and SNAPSHOT window config types. TYPE: |
training_dataset_version | int, optional The version of the training dataset to use as reference. Only used for TRAINING_DATASET window config type. TYPE: |
specific_value | float, optional The specific value to use as reference. Only used for SPECIFIC_VALUE window config type. TYPE: |
row_percentage | float, optional The fraction of rows to use when computing statistics [0, 1.0]. Only used for ROLLING_TIME and ALL_TIME window config types. TYPE: |
| RAISES | DESCRIPTION |
|---|---|
AttributeError | If window_config_type is not one of INSERT, SNAPSHOT, BATCH, TRAINING_DATASET, SPECIFIC_VALUE. |