ism3d.modeling.model.model_setup

ism3d.modeling.model.model_setup(mod_dct, dat_dct, verbose=False)

create model container

this function can be ran only once before starting fitting iteration, so that the memory allocation/ allication will happen once during a fitting run. The output will provide the dataframework where the reference model can be mapped into.

it will also initilize some informationed used for modeling (e.g. sampling array / header / WCS)

notes on evaluating efficiency:

While building the intrinsic data-model from a physical model can be expensive, the simulated observation (2D/3D convolution) is usually the bottle-neck.

some tips to improve the effeciency:

• exclude empty (masked/flux=0) region for the convolution

• joint all objects in the intrinsic model before the convolution, e.g.

  overlapping objects, lines

• use to low-dimension convolution when possible (e.g. for the narrow-band continumm)

before splitting line & cont models:

— apicall : 2.10178 seconds —

after splitting line & cont models:

— apicall : 0.84662 seconds —

note: imod2dHold emission componnets with Frequency-Dependent Spatial Distribution

imod3d : Hold emission conponents with Frequency-Dependent Spatial Distribution imodel=imod2d+imod3d : We always keep a copy of imod2d and imod3d to improve the effeicnecy in simobs()

uvmodel: np.complex64

imodel: np.float32