probe from numpy array will not be rescaled to photon count #596
Conversation
Patch release 0.7.1
Release 0.8
Patch release 0.8.1
Patch Release 0.8.2
Bump version to 0.8.2 (#575)
|
Currently the options for illumination.photons are to type None (default) or a number of photons. This feature however works only for a probe that is generated from a model and not for probes initialized from a .ptyr file or a numpy array. With those two options (None or a number) it is not possible to clearly extend the scaling functionality to probes initialized from a .ptyr file or a numpy array. I suggest a new accepted string for illumination.photons (something like "maxdiff") to specifically choose to
It is fairly simple to implement and would unify the features functionality across all methods of loading / creating the initial probe. With "maxdiff" as default any old script that did not specifically set illumination.photons to None to achieve the scaling to the brightest diffraction pattern would still work as before. Only scripts that for some reason did set illumination.photons=None would change their behavior, as in the suggested case, no more scaling would be applied. Any opinions from the ptypy guardians are welcome. |
|
python 3.8 tests fail when installing dependencies... not in the changed bits of the code |
Tests should work if you create the PR against dev. Should also male sure your branch is rebased against dev. |
* Drop Python <= 3.8 support * Add py 3.13 in CI tests * Use conda-incubator to set up conda env * Ensure correct activation throughout the tests * Migrate to importlib in script2rst * Migrate to importlib in resources * Migrate to packaging.version
* Include accelerate base tests by default * Change expected output in FFT filter tests to reflect support for single precision in Numpy > 2.0 * Adjust tolerance in pycuda fft filter tests
When a probe was initalized from a numpy array, the _process(ing) applied to that loaded wavefront always scaled it to the number of photons estimated from the loaded diffraction patterns. This happened even when illumiantion.photons was specifically set to None:
p.scans.scan00.illumination.photons = NoneA similar bug was fixed when loading probes from previous reconstruction (.ptyr) files by specifically setting p.photons to None again. The same "fix" can be used here as well.
The loaded numpy array now remains unchanged.
But now it is also impossible to scale a loaded np.array if the users specifically wishes for it.