Dataset And Sampler Reference#

Dataset entries are used under data.boundaries, data.validation, data.sampler, and data.samplers.

Dataset Types#

Type

Role

Description

analytical

benchmark boundary

Low & Lou analytical fields for case1/case2 benchmarks.

sharp

cartesian boundary

SHARP CEA Br/Bt/Bp FITS input, optionally with uncertainty maps.

fits

cartesian boundary

Generic Cartesian Br/Bt/Bp FITS files.

los_trv_azi

cartesian boundary

LOS/transverse/azimuth FITS input with automatic disambiguation support.

los

cartesian boundary

Line-of-sight-only Cartesian boundary input.

fld_inc_azi

cartesian boundary

Field strength/inclination/azimuth FITS input.

height

cartesian sampler

Random Cartesian volume samples for force-free and potential losses.

potential

cartesian boundary

Potential-field boundary generated from the photospheric boundary.

map

spherical boundary

Spherical Br/Bt/Bp map input for full-disk or synoptic maps.

random_radial_grouped

spherical sampler

Grouped radial samples in the spherical volume.

sphere

spherical validation

Spherical validation grid for field-quality callbacks.

spherical_slices

spherical validation

Radial or angular slices for spherical visualization callbacks.

analytical#

Role: benchmark boundary.

Key

Type

Default

Description

case

int | str

1

Analytical Low & Lou case identifier.

bounds

list[float]

loader default

Physical Cartesian volume bounds for the benchmark field.

resolution

int | list[int]

loader default

Sampling resolution used to generate the benchmark boundary.

batch_size

int

shared boundary batch size

Boundary samples per training batch.

sharp#

Role: cartesian boundary or validation.

Key

Type

Default

Description

files.Br

str

required

Radial magnetic-field FITS file.

files.Bt

str

required

Theta/transverse magnetic-field FITS file.

files.Bp

str

required

Phi/azimuthal magnetic-field FITS file.

errors.*

str

optional

Optional uncertainty FITS files, stored as error_path internally.

Mm_per_pixel

float

FITS/WCS default

Boundary plate scale in Mm per pixel.

coordinate_center

list[float] | dict

[0, 0]

Cartesian coordinate of the image center in Mm.

slice

slice | null

null

Optional image subset.

batch_size

int

shared boundary batch size

Boundary samples per training or validation batch.

fits#

Role: cartesian boundary or validation.

Key

Type

Default

Description

files.Br

str

required

Radial/vertical magnetic-field FITS file.

files.Bt

str

required

Transverse x/theta component FITS file.

files.Bp

str

required

Transverse y/phi component FITS file.

Mm_per_pixel

float

dataset default

Boundary plate scale in Mm per pixel.

coordinate_center

list[float] | dict

[0, 0]

Cartesian coordinate of the image center in Mm.

batch_size

int

shared boundary batch size

Boundary samples per training or validation batch.

los_trv_azi#

Role: cartesian boundary or validation.

Key

Type

Default

Description

files.B_los

str

required

Line-of-sight magnetic-field FITS file.

files.B_trv

str

required

Transverse-field magnitude FITS file.

files.B_azi

str

required

Azimuth FITS file.

ambiguous_azimuth

bool

false

Train/evaluate both azimuth disambiguation branches when supported by the loss.

load_map

bool

dataset default

Load map metadata for coordinate conversion and plotting.

height_mapping

dict

none

Map this boundary to a height or learned height range, for multi-height runs.

Mm_per_pixel

float

dataset default

Boundary plate scale in Mm per pixel.

coordinate_center

list[float] | dict

[0, 0]

Cartesian coordinate of the image center in Mm.

batch_size

int

shared boundary batch size

Boundary samples per training or validation batch.

los#

Role: cartesian boundary or validation.

Key

Type

Default

Description

files.B_los

str

required

Line-of-sight magnetic-field FITS file.

Mm_per_pixel

float

dataset default

Boundary plate scale in Mm per pixel.

coordinate_center

list[float] | dict

[0, 0]

Cartesian coordinate of the image center in Mm.

batch_size

int

shared boundary batch size

Boundary samples per training or validation batch.

fld_inc_azi#

Role: cartesian boundary or validation.

Key

Type

Default

Description

files.B

str

required

Field-strength FITS file.

files.inc

str

required

Inclination FITS file.

files.azi

str

required

Azimuth FITS file.

Mm_per_pixel

float

dataset default

Boundary plate scale in Mm per pixel.

coordinate_center

list[float] | dict

[0, 0]

Cartesian coordinate of the image center in Mm.

batch_size

int

shared boundary batch size

Boundary samples per training or validation batch.

height#

Role: cartesian sampler.

Key

Type

Default

Description

batch_size

int

16384 in examples

Random volume samples per training batch.

z_sample

int

128

Number of heights sampled per xy location.

z_sampling_exponent

float

2

Bias random samples toward lower heights when greater than 1.

length

int | null

data.iterations

Number of sampler batches per epoch-like pass.

requires_jacobian

bool

true

Required for force-free, divergence, and potential losses.

potential#

Role: cartesian potential boundary.

Key

Type

Default

Description

type

potential | potential_top | none

potential

Potential-field side/top boundary mode.

strides

int

4

Downsample factor for generating the potential boundary.

method

fft | direct

fft

Potential-field solver.

batch_size

int

boundary batch size / 4

Potential boundary samples per batch.

requires_jacobian

bool

false

Usually disabled for boundary matching.

cube#

Role: cartesian validation.

Key

Type

Default

Description

ds_per_pixel

float

1 / validation_pixel_per_ds

Validation grid spacing in model units.

batch_size

int

validation_batch_size

Validation samples per batch.

requires_jacobian

bool

true

Needed for quality metrics that use derivatives.

slices#

Role: cartesian validation.

Key

Type

Default

Description

n_slices

int

10

Number of height slices to evaluate and plot.

batch_size

int

4096

Validation samples per batch.

requires_jacobian

bool

true

Needed for current-density plots.

map#

Role: spherical boundary or validation.

Key

Type

Default

Description

files.Br

str

required

Radial magnetic-field map.

files.Bt

str

required

Theta magnetic-field map.

files.Bp

str

required

Phi magnetic-field map.

errors.Br_err

str

optional

Radial uncertainty map, merged into the file map internally.

errors.Bt_err

str

optional

Theta uncertainty map, merged into the file map internally.

errors.Bp_err

str

optional

Phi uncertainty map, merged into the file map internally.

mask_configs

dict

optional

Optional masking/filtering, for example a mu_filter.

batch_size

int

data.batch_size or dataset default

Map samples per batch.

requires_jacobian

bool

false for boundary examples

Enable only when the dataset feeds derivative-based losses.

random_radial_grouped#

Role: spherical sampler.

Key

Type

Default

Description

batch_size

int

16384 in examples

Random spherical volume samples per batch.

n_lat_lon_sample

int

128

Number of angular samples; batch_size must be divisible by this value.

radial_sampling_exponent

float

1

Bias radial samples when greater than 1.

latitude_range

list[float]

[-90, 90]

Latitude range in degrees unless unit is changed.

longitude_range

list[float]

[0, 360]

Longitude range in degrees unless unit is changed.

length

int | null

data.iterations

Internal sampler length; public YAML should usually use data.iterations.

requires_jacobian

bool

true

Required for force-free, potential, and energy-gradient losses.

sphere#

Role: spherical validation.

Key

Type

Default

Description

resolution

int

256

Radial and longitude resolution of the validation sphere.

batch_size

int

1024

Validation samples per batch.

latitude_range

list[float]

[-90, 90]

Latitude range in degrees unless unit is changed.

longitude_range

list[float]

[0, 360]

Longitude range in degrees unless unit is changed.

requires_jacobian

bool

true

Needed for derivative-based quality metrics.

spherical_slices#

Role: spherical validation.

Key

Type

Default

Description

longitude_resolution

int

256

Longitude resolution for slice plots.

n_slices

int

5

Number of radial slices.

batch_size

int

1024

Validation samples per batch.

latitude_range

list[float]

[-90, 90]

Latitude range in degrees unless unit is changed.

longitude_range

list[float]

[0, 360]

Longitude range in degrees unless unit is changed.

requires_jacobian

bool

true

Needed for current-density plots.