CalcZernikesNeuralTaskConfig#

class lsst.ts.wep.task.CalcZernikesNeuralTaskConfig(*args, **kw)#

Bases: CalcZernikesTaskConfig

Configuration for CalcZernikesNeuralTask.

wavenetPath#

Model weights path for wavenet. If None, TARTS will create a new model with random weights (useful for testing).

Type:: str or None

alignetPath#

Model weights path for alignet. If None, TARTS will create a new model with random weights (useful for testing).

Type:: str or None

aggregatornetPath#

Model weights path for aggregatornet. If None, TARTS will create a new model with random weights (useful for testing).

Type:: str or None

oodModelPath#

Directory path for the OOD model to be used by TARTS for out-of- distribution detection. If None, OOD checks are disabled.

Type:: str or None

datasetParamPath#

Path to TARTS dataset parameters YAML file containing normalization scaling factors, image processing parameters (CROP_SIZE, deg_per_pix, mm_pix), model hyperparameters, and training data file paths.

Type:: str

device#

Device to use for calculations. Options: ‘cpu’ for CPU computation, ‘cuda’ for GPU computation (default). Any value other than ‘cpu’ will use CUDA.

Type:: str

nollIndices#

List of Noll indices to calculate. Default is Z4-Z22 (4-22), excluding piston (Z1), tip (Z2), and tilt (Z3) which are typically not measured in wavefront sensing. Values must be unique, ascending, at least 4, and include complete azimuthal pairs (same rules as EstimateZernikesBaseTask; enforced by checkNollIndices at task initialization).

Type:: list[int]

cropSize#

Size of donut crop in pixels (width and height). Default is 160 pixels, which matches the TARTS neural network training data format.

Type:: int

intraDfcDist#

Defocal distance for intra-focal images in mm. Negative value indicates inward defocus. Default is -1.5 mm.

Type:: float

extraDfcDist#

Defocal distance for extra-focal images in mm. Positive value indicates outward defocus. Default is 1.5 mm.

Type:: float

Attributes Summary

`aggregatornetPath`	Model Weights Path for aggregatornet (`str`, default `None`)
`alignetPath`	Model Weights Path for alignet (`str`, default `None`)
`combineZernikes`	Choice of task to combine the Zernikes from pairs of donuts into a single value for the detector.
`connections`	Field which refers to a dynamically added configuration class which is based on a PipelineTaskConnections class.
`cropSize`	Size of donut crop in pixels (width and height).
`datasetParamPath`	Path to TARTS dataset parameters YAML file containing normalization scaling factors, image processing parameters (CROP_SIZE, deg_per_pix, mm_pix), model hyperparameters, and training data file paths (`str`)
`device`	'cpu' for CPU computation, 'cuda' for GPU computation (default).
`doBlurClip`	Remove donuts with outlier donut blur fwhm fromfinal averages.
`doDonutStampSelector`	Whether or not to run donut stamp selector.If this is False, then we do not get donutQualityTable.(The default is True).
`donutStampSelector`	How to select donut stamps.
`estimateZernikes`	Choice of task to estimate Zernikes from pairs of donuts.
`extraDfcDist`	Defocal distance for extra-focal images in mm.
`intraDfcDist`	Defocal distance for intra-focal images in mm.
`nollIndices`	List of Noll indices to calculate.
`oodModelPath`	Directory path for OOD model used by TARTS (optional) (`str`, default `None`)
`saveLogOutput`	Flag to enable/disable saving of log output for a task, enabled by default.
`wavenetPath`	Model Weights Path for wavenet (`str`, default `None`)

Attributes Documentation

aggregatornetPath: Field#: Model Weights Path for aggregatornet (str, default None)

alignetPath: Field#: Model Weights Path for alignet (str, default None)

combineZernikes: pexConfig.ConfigurableField#: Choice of task to combine the Zernikes from pairs of donuts into a single value for the detector. (The default is CombineZernikesSigmaClipTask.) (ConfigurableInstance, default <class 'lsst.ts.wep.task.combineZernikesSigmaClipTask.CombineZernikesSigmaClipTaskConfig'>)

connections: pexConfig.ConfigField#: Field which refers to a dynamically added configuration class which is based on a PipelineTaskConnections class.

cropSize: Field#: Size of donut crop in pixels (width and height). Default is 160 pixels, which matches the TARTS neural network training data format. (int, default 160)

datasetParamPath: Field#: Path to TARTS dataset parameters YAML file containing normalization scaling factors, image processing parameters (CROP_SIZE, deg_per_pix, mm_pix), model hyperparameters, and training data file paths (str)

device: Field#

‘cpu’ for CPU computation, ‘cuda’ for GPU computation (default). Any value other than ‘cpu’ will use CUDA. (str, default 'cuda')

Type:: Device to use for calculations. Options

doBlurClip: pexConfig.Field#: Remove donuts with outlier donut blur fwhm fromfinal averages. (bool, default True)

doDonutStampSelector: pexConfig.Field#: Whether or not to run donut stamp selector.If this is False, then we do not get donutQualityTable.(The default is True). It is also possible to rundonut stamp selector (with this config set to True), butturn off doSelection config inside the donut stamp selector,which would return all donuts as selected, as well asreturning a quality table. (bool, default True)

donutStampSelector: pexConfig.ConfigurableField#: How to select donut stamps. (ConfigurableInstance, default <class 'lsst.ts.wep.task.donutStampSelectorTask.DonutStampSelectorTaskConfig'>)

estimateZernikes: pexConfig.ConfigurableField#: Choice of task to estimate Zernikes from pairs of donuts. (the default is EstimateZernikesTieTask) (ConfigurableInstance, default <class 'lsst.ts.wep.task.estimateZernikesDanishTask.EstimateZernikesDanishConfig'>)

extraDfcDist: Field#: Defocal distance for extra-focal images in mm. Positive value indicates outward defocus. (float, default 1.5)

intraDfcDist: Field#: Defocal distance for intra-focal images in mm. Negative value indicates inward defocus. (float, default -1.5)

nollIndices: Field#: List of Noll indices to calculate. Default is Z4-Z22 (4-22), excluding piston (Z1), tip (Z2), and tilt (Z3) which are typically not measured in wavefront sensing. Indices must be unique, ascending, >= 4, and include complete azimuthal pairs (see checkNollIndices in zernikeUtils). (List, default [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22])

oodModelPath: Field#: Directory path for OOD model used by TARTS (optional) (str, default None)

saveLogOutput#: Flag to enable/disable saving of log output for a task, enabled by default. (bool, default True)

wavenetPath: Field#: Model Weights Path for wavenet (str, default None)