SimpleDiskEnvFit issueshttps://gitlab.mpcdf.mpg.de/szucs/SimpleDiskEnvFit/-/issues2019-08-14T11:46:08Zhttps://gitlab.mpcdf.mpg.de/szucs/SimpleDiskEnvFit/-/issues/24Combined MPI - OpenMP multiprocessing2019-08-14T11:46:08ZLaszlo SzucsCombined MPI - OpenMP multiprocessingWhen less walkers are initialized than cores available, then some of the cores idle. In order to take advantage of them, call the following tasks in parallel mode:
- thermal MC
- image computation
- galarioWhen less walkers are initialized than cores available, then some of the cores idle. In order to take advantage of them, call the following tasks in parallel mode:
- thermal MC
- image computation
- galariohttps://gitlab.mpcdf.mpg.de/szucs/SimpleDiskEnvFit/-/issues/13Memory management2019-08-14T09:47:20ZLaszlo SzucsMemory managementMinimize memory footprint
-------------------------
To free memory use:
```python
import gc
del u1, u2, u3, ... # 0.6% -> 700 MB
gc.collect() # 0.2% -> 250 MB
```
Note that del only deletes references to object. If no referen...Minimize memory footprint
-------------------------
To free memory use:
```python
import gc
del u1, u2, u3, ... # 0.6% -> 700 MB
gc.collect() # 0.2% -> 250 MB
```
Note that del only deletes references to object. If no references are associated with object, garbage collection deletes object. However, if other references exists, then object remains in memory.https://gitlab.mpcdf.mpg.de/szucs/SimpleDiskEnvFit/-/issues/12Speed up code2019-08-14T09:07:15ZLaszlo SzucsSpeed up codeWhen multiple wavelength ALMA data is fitted, then the per model computation time is large. Find out which modelling steps are the most time consuming and optimize as much as possible.
Possible optimization:
- do not compute and store s...When multiple wavelength ALMA data is fitted, then the per model computation time is large. Find out which modelling steps are the most time consuming and optimize as much as possible.
Possible optimization:
- do not compute and store synthetic visibility when not needed
- try ifast = 1, 2 in the thermal Monte Carlo of RADMC-3D
- compute synthetic images at all wavelength simultaneously (possible if scattering is off)
- print timing information on requestv0.2https://gitlab.mpcdf.mpg.de/szucs/SimpleDiskEnvFit/-/issues/11Create synthetic test2019-05-09T08:32:55ZLaszlo SzucsCreate synthetic testCreate synthetic visibilities using RADMC-3D and fit them using SimpleDiskEnvFit to recover the model parameters.Create synthetic visibilities using RADMC-3D and fit them using SimpleDiskEnvFit to recover the model parameters.v0.2https://gitlab.mpcdf.mpg.de/szucs/SimpleDiskEnvFit/-/issues/5Add option to provide photometric observational constraint2019-03-27T11:03:59ZLaszlo SzucsAdd option to provide photometric observational constraintradmc3dRunner should get a new function to compute SED at user specified wavelengths (in the future also to compute flux in band instead of single wavelength). \chi^2 is computed from the SED and the supplied observational flux and added...radmc3dRunner should get a new function to compute SED at user specified wavelengths (in the future also to compute flux in band instead of single wavelength). \chi^2 is computed from the SED and the supplied observational flux and added to the total model \chi^2 used in the MCMC algorithm. The \chi^2 should be weighted, because otherwise the visibility data would dominate it.v0.2https://gitlab.mpcdf.mpg.de/szucs/SimpleDiskEnvFit/-/issues/1Add option to save computed images2019-07-11T15:17:04ZLaszlo SzucsAdd option to save computed imagesUse HDF5 to store the data. Advantage is that independent runners may write the output file parallel.
In run_mcmc() function initialize and open HDF5 file, pass handler / ID / object to emcee runners. Store / save data in lnpost() funct...Use HDF5 to store the data. Advantage is that independent runners may write the output file parallel.
In run_mcmc() function initialize and open HDF5 file, pass handler / ID / object to emcee runners. Store / save data in lnpost() function.v0.2