ScarletDeblendConfig¶
- class lsst.meas.extensions.scarlet.ScarletDeblendConfig(*args, **kw)¶
Bases:
Config
MultibandDeblendConfig
Configuration for the multiband deblender. The parameters are organized by the parameter types, which are - Stopping Criteria: Used to determine if the fit has converged - Position Fitting Criteria: Used to fit the positions of the peaks - Constraints: Used to apply constraints to the peaks and their components - Other: Parameters that don’t fit into the above categories
Attributes Summary
Fraction of background to use for a sparsity threshold.
Whether or not to process isolated sources in the deblender (
List
, default['BAD', 'CR', 'NO_DATA', 'SAT', 'SUSPECT', 'EDGE']
)If True, catch exceptions thrown by the deblender, log them, and set a flag on the parent, instead of letting them propagate up (
bool
, defaultTrue
)Only deblend parent Footprints with a number of peaks in the (inclusive) range indicated.If
useCiLimits==False
then this parameter is ignored.Only use the first
ciNumParentsToDeblend
parent footprints with a total peak count withinciDebledChildRange
.Columns to pass from the parent to the child.
Type of convolution to render the model to the observations.
Whether or not to fallback to a smaller number of components if a source does not initialize (
bool
, defaultTrue
)Mask planes with the corresponding limit on the fraction of masked pixels.
Maximum rectangular footprint area * nPeaks in the footprint.
Maximum area for footprints before they are ignored as large; non-positive means no threshold applied (
int
, default100000
)Maximum linear dimension for footprints before they are ignored as large; non-positive means no threshold applied (
int
, default0
)Maximum number of iterations to deblend a single parent (
int
, default300
)Only deblend the brightest maxNumberOfPeaks peaks in the parent (<= 0: unlimited) (
int
, default200
)Maximum number of proximal operator iterations inside of each iteration of the optimizer.
Maximum number of pixels * number of sources in a blend.
Minimum axis ratio for footprints before they are ignored as large; non-positive means no threshold applied (
float
, default0.0
)Minimum number of iterations before the optimizer is allowed to stop.
Minimum Signal to noise to accept the source.Sources with lower flux will be initialized with the PSF but updated like an ordinary ExtendedSource (known in scarlet as a
CompactSource
).Define sigma for the model frame PSF (
float
, default0.8
)Model PSF side length in pixels (
int
, default11
)The type of image to use for initializing the morphology.
Fraction of background RMS a pixel must haveto be included in the initial morphology (
float
, default1
)Mask name for footprints not deblended, or None (
str
, default'NOT_DEBLENDED'
)The optimizer to use for fitting parameters and is only used when version='lite' (
str
, default'adaprox'
)Whether or not to process isolated sources in the deblender (
bool
, defaultTrue
)Names of flags which should never be deblended.
Change in the loss function between iterations to exit fitter.
Whether or not to save the SEDs and templates (
bool
, defaultTrue
)Whether or not to solve for the best-fit spectra during initialization.
How to determine which model to use for sources, from - 'single': use a single component for all sources - 'double': use a bulge disk model for all sources - 'compact': use a single component model, initialzed with a point source morphology, for all sources - 'point': use a point-source model for all sources - 'fit: use a PSF fitting model to determine the number of components (not yet implemented) Deprecated: This field will be deprecated when the default for
version
is changed tolite
.Mask planes to ignore when performing statistics (
List
, default['SAT', 'INTRP', 'NO_DATA']
)Limit the number of sources deblended for CI to prevent long build times (
bool
, defaultFalse
)Whether or not use use inverse variance weighting.If
useWeights
isFalse
then flat weights are used (bool
, defaultTrue
)The version of scarlet to use.
Number of wavelet scales to use for wavelet initialization.
Methods Summary
compare
(other[, shortcut, rtol, atol, output])Compare this configuration to another
Config
for equality.formatHistory
(name, **kwargs)Format a configuration field's history to a human-readable format.
freeze
()Make this config, and all subconfigs, read-only.
items
()Get configurations as
(field name, field value)
pairs.keys
()Get field names.
load
(filename[, root])Modify this config in place by executing the Python code in a configuration file.
loadFromStream
(stream[, root, filename, ...])Modify this Config in place by executing the Python code in the provided stream.
loadFromString
(code[, root, filename, ...])Modify this Config in place by executing the Python code in the provided string.
names
()Get all the field names in the config, recursively.
save
(filename[, root])Save a Python script to the named file, which, when loaded, reproduces this config.
saveToStream
(outfile[, root, skipImports])Save a configuration file to a stream, which, when loaded, reproduces this config.
saveToString
([skipImports])Return the Python script form of this configuration as an executable string.
Subclass hook for computing defaults.
toDict
()Make a dictionary of field names and their values.
update
(**kw)Update values of fields specified by the keyword arguments.
validate
()Validate the Config, raising an exception if invalid.
values
()Get field values.
Attributes Documentation
- backgroundThresh¶
Fraction of background to use for a sparsity threshold. This prevents sources from growing unrealistically outside the parent footprint while still modeling flux correctly for bright sources. (
float
, default0.25
)
- badMask¶
Whether or not to process isolated sources in the deblender (
List
, default['BAD', 'CR', 'NO_DATA', 'SAT', 'SUSPECT', 'EDGE']
)
- catchFailures¶
If True, catch exceptions thrown by the deblender, log them, and set a flag on the parent, instead of letting them propagate up (
bool
, defaultTrue
)
- ciDeblendChildRange¶
Only deblend parent Footprints with a number of peaks in the (inclusive) range indicated.If
useCiLimits==False
then this parameter is ignored. (List
, default[5, 10]
)
- ciNumParentsToDeblend¶
Only use the first
ciNumParentsToDeblend
parent footprints with a total peak count withinciDebledChildRange
. IfuseCiLimits==False
then this parameter is ignored. (int
, default10
)
- columnInheritance¶
Columns to pass from the parent to the child. The key is the name of the column for the parent record, the value is the name of the column to use for the child. (
Dict
, default{'deblend_nChild': 'deblend_parentNChild', 'deblend_nPeaks': 'deblend_parentNPeaks', 'deblend_spectrumInitFlag': 'deblend_spectrumInitFlag', 'deblend_blendConvergenceFailedFlag': 'deblend_blendConvergenceFailedFlag'}
)
- convolutionType¶
Type of convolution to render the model to the observations. - ‘fft’: perform convolutions in Fourier space - ‘real’: peform convolutions in real space. (
str
, default'fft'
)
- fallback¶
Whether or not to fallback to a smaller number of components if a source does not initialize (
bool
, defaultTrue
)
- history¶
Read-only history.
- maskLimits¶
Mask planes with the corresponding limit on the fraction of masked pixels. Sources violating this limit will not be deblended. If the fraction is
0
then the limit is a single pixel. (Dict
, default{}
)
- maxAreaTimesPeaks¶
Maximum rectangular footprint area * nPeaks in the footprint. This was introduced in DM-33690 to prevent fields that are crowded or have a LSB galaxy that causes memory intensive initialization in scarlet from dominating the overall runtime and/or causing the task to run out of memory. (<= 0: unlimited) (
int
, default10000000
)
- maxFootprintArea¶
Maximum area for footprints before they are ignored as large; non-positive means no threshold applied (
int
, default100000
)
- maxFootprintSize¶
Maximum linear dimension for footprints before they are ignored as large; non-positive means no threshold applied (
int
, default0
)
- maxNumberOfPeaks¶
Only deblend the brightest maxNumberOfPeaks peaks in the parent (<= 0: unlimited) (
int
, default200
)
- maxProxIter¶
Maximum number of proximal operator iterations inside of each iteration of the optimizer. This config field is only used if version=’lite’ and optimizer=’adaprox’. (
int
, default1
)
- maxSpectrumCutoff¶
Maximum number of pixels * number of sources in a blend. This is different than
maxFootprintArea
because this isn’t the footprint area but the area of the bounding box that contains the footprint, and is also multiplied by the number ofsources in the footprint. This prevents large skinny blends with a high density of sources from running out of memory. IfmaxSpectrumCutoff == -1
then there is no cutoff. (int
, default1000000
)
- minFootprintAxisRatio¶
Minimum axis ratio for footprints before they are ignored as large; non-positive means no threshold applied (
float
, default0.0
)
- minSNR¶
Minimum Signal to noise to accept the source.Sources with lower flux will be initialized with the PSF but updated like an ordinary ExtendedSource (known in scarlet as a
CompactSource
). (float
, default50
)
- morphImage¶
The type of image to use for initializing the morphology. Must be either ‘chi2’ or ‘wavelet’. (
str
, default'chi2'
)Allowed values:
'chi2'
Initialize sources on a chi^2 image made from all available bands
'wavelet'
Initialize sources using a wavelet decomposition of the chi^2 image
'None'
Field is optional
- morphThresh¶
Fraction of background RMS a pixel must haveto be included in the initial morphology (
float
, default1
)
- optimizer¶
The optimizer to use for fitting parameters and is only used when version=’lite’ (
str
, default'adaprox'
)Allowed values:
'adaprox'
Proximal ADAM optimization
'fista'
Accelerated proximal gradient method
'None'
Field is optional
- pseudoColumns¶
Names of flags which should never be deblended. (
List
, default['merge_peak_sky', 'sky_source']
)
- relativeError¶
Change in the loss function between iterations to exit fitter. Typically this is
1e-2
if measurements will be made on the flux re-distributed models and1e-4
when making measurements on the models themselves. (float
, default0.01
)
- setSpectra¶
Whether or not to solve for the best-fit spectra during initialization. This makes initialization slightly longer, as it requires a convolution to set the optimal spectra, but results in a much better initial log-likelihood and reduced total runtime, with convergence in fewer iterations.This option is only used when peaks*area <
maxSpectrumCutoff
will use the improved initialization. (bool
, defaultTrue
)
- sourceModel¶
How to determine which model to use for sources, from - ‘single’: use a single component for all sources - ‘double’: use a bulge disk model for all sources - ‘compact’: use a single component model, initialzed with a point source morphology, for all sources - ‘point’: use a point-source model for all sources - ‘fit: use a PSF fitting model to determine the number of components (not yet implemented) Deprecated: This field will be deprecated when the default for
version
is changed tolite
. (str
, default'double'
)
- statsMask¶
Mask planes to ignore when performing statistics (
List
, default['SAT', 'INTRP', 'NO_DATA']
)
- useCiLimits¶
Limit the number of sources deblended for CI to prevent long build times (
bool
, defaultFalse
)
- useWeights¶
Whether or not use use inverse variance weighting.If
useWeights
isFalse
then flat weights are used (bool
, defaultTrue
)
- version¶
The version of scarlet to use. (
str
, default'lite'
)Allowed values:
'old_lite'
scarlet lite from the scarlet main package
'lite'
LSST optimized version of scarlet for survey data from a single instrument
'None'
Field is optional
- waveletScales¶
Number of wavelet scales to use for wavelet initialization. This field is only used when
version`='lite' and `morphImage`='wavelet'. (`int
, default5
)
Methods Documentation
- compare(other, shortcut=True, rtol=1e-08, atol=1e-08, output=None)¶
Compare this configuration to another
Config
for equality.- Parameters:
- other
lsst.pex.config.Config
Other
Config
object to compare against this config.- shortcut
bool
, optional If
True
, return as soon as an inequality is found. Default isTrue
.- rtol
float
, optional Relative tolerance for floating point comparisons.
- atol
float
, optional Absolute tolerance for floating point comparisons.
- outputcallable, optional
A callable that takes a string, used (possibly repeatedly) to report inequalities.
- other
- Returns:
- isEqual
bool
True
when the twolsst.pex.config.Config
instances are equal.False
if there is an inequality.
- isEqual
See also
Notes
Unselected targets of
RegistryField
fields and unselected choices ofConfigChoiceField
fields are not considered by this method.Floating point comparisons are performed by
numpy.allclose
.
- formatHistory(name, **kwargs)¶
Format a configuration field’s history to a human-readable format.
- Parameters:
- name
str
Name of a
Field
in this config.- **kwargs
Keyword arguments passed to
lsst.pex.config.history.format
.
- name
- Returns:
- history
str
A string containing the formatted history.
- history
See also
- freeze()¶
Make this config, and all subconfigs, read-only.
- items()¶
Get configurations as
(field name, field value)
pairs.- Returns:
- items
ItemsView
Iterator of tuples for each configuration. Tuple items are:
Field name.
Field value.
- items
- keys()¶
Get field names.
- Returns:
- names
KeysView
List of
lsst.pex.config.Field
names.
- names
- load(filename, root='config')¶
Modify this config in place by executing the Python code in a configuration file.
- Parameters:
- filename
str
Name of the configuration file. A configuration file is Python module.
- root
str
, optional Name of the variable in file that refers to the config being overridden.
For example, the value of root is
"config"
and the file contains:config.myField = 5
Then this config’s field
myField
is set to5
.
- filename
- loadFromStream(stream, root='config', filename=None, extraLocals=None)¶
Modify this Config in place by executing the Python code in the provided stream.
- Parameters:
- streamfile-like object,
str
,bytes
, orCodeType
Stream containing configuration override code. If this is a code object, it should be compiled with
mode="exec"
.- root
str
, optional Name of the variable in file that refers to the config being overridden.
For example, the value of root is
"config"
and the file contains:config.myField = 5
Then this config’s field
myField
is set to5
.- filename
str
, optional Name of the configuration file, or
None
if unknown or contained in the stream. Used for error reporting.- extraLocals
dict
ofstr
toobject
, optional Any extra variables to include in local scope when loading.
- streamfile-like object,
See also
Notes
For backwards compatibility reasons, this method accepts strings, bytes and code objects as well as file-like objects. New code should use
loadFromString
instead for most of these types.
- loadFromString(code, root='config', filename=None, extraLocals=None)¶
Modify this Config in place by executing the Python code in the provided string.
- Parameters:
- code
str
,bytes
, orCodeType
Stream containing configuration override code.
- root
str
, optional Name of the variable in file that refers to the config being overridden.
For example, the value of root is
"config"
and the file contains:config.myField = 5
Then this config’s field
myField
is set to5
.- filename
str
, optional Name of the configuration file, or
None
if unknown or contained in the stream. Used for error reporting.- extraLocals
dict
ofstr
toobject
, optional Any extra variables to include in local scope when loading.
- code
- Raises:
- ValueError
Raised if a key in extraLocals is the same value as the value of the root argument.
- save(filename, root='config')¶
Save a Python script to the named file, which, when loaded, reproduces this config.
- Parameters:
- filename
str
Desination filename of this configuration.
- root
str
, optional Name to use for the root config variable. The same value must be used when loading (see
lsst.pex.config.Config.load
).
- filename
- saveToStream(outfile, root='config', skipImports=False)¶
Save a configuration file to a stream, which, when loaded, reproduces this config.
- Parameters:
- outfilefile-like object
Destination file object write the config into. Accepts strings not bytes.
- root
str
, optional Name to use for the root config variable. The same value must be used when loading (see
lsst.pex.config.Config.load
).- skipImports
bool
, optional If
True
then do not includeimport
statements in output, this is to support human-oriented output frompipetask
where additional clutter is not useful.
- saveToString(skipImports=False)¶
Return the Python script form of this configuration as an executable string.
- Parameters:
- Returns:
- code
str
A code string readable by
loadFromString
.
- code
- setDefaults()¶
Subclass hook for computing defaults.
Notes
Derived
Config
classes that must compute defaults rather than using theField
instances’s defaults should do so here. To correctly use inherited defaults, implementations ofsetDefaults
must call their base class’ssetDefaults
.
- toDict()¶
Make a dictionary of field names and their values.
See also
Notes
This method uses the
toDict
method of individual fields. Subclasses ofField
may need to implement atoDict
method for this method to work.
- update(**kw)¶
Update values of fields specified by the keyword arguments.
- Parameters:
- **kw
Keywords are configuration field names. Values are configuration field values.
Notes
The
__at
and__label
keyword arguments are special internal keywords. They are used to strip out any internal steps from the history tracebacks of the config. Do not modify these keywords to subvert aConfig
instance’s history.Examples
This is a config with three fields:
>>> from lsst.pex.config import Config, Field >>> class DemoConfig(Config): ... fieldA = Field(doc='Field A', dtype=int, default=42) ... fieldB = Field(doc='Field B', dtype=bool, default=True) ... fieldC = Field(doc='Field C', dtype=str, default='Hello world') ... >>> config = DemoConfig()
These are the default values of each field:
>>> for name, value in config.iteritems(): ... print(f"{name}: {value}") ... fieldA: 42 fieldB: True fieldC: 'Hello world'
Using this method to update
fieldA
andfieldC
:>>> config.update(fieldA=13, fieldC='Updated!')
Now the values of each field are:
>>> for name, value in config.iteritems(): ... print(f"{name}: {value}") ... fieldA: 13 fieldB: True fieldC: 'Updated!'
- validate()¶
Validate the Config, raising an exception if invalid.
- Raises:
- lsst.pex.config.FieldValidationError
Raised if verification fails.
Notes
The base class implementation performs type checks on all fields by calling their
validate
methods.Complex single-field validation can be defined by deriving new Field types. For convenience, some derived
lsst.pex.config.Field
-types (ConfigField
andConfigChoiceField
) are defined inlsst.pex.config
that handle recursing into subconfigs.Inter-field relationships should only be checked in derived
Config
classes after calling this method, and base validation is complete.
- values()¶
Get field values.
- Returns:
- values
ValuesView
Iterator of field values.
- values