refactor: cache image_response.contents in image_deconvolution (#3)

* hidden: profiling nb and removal of profiling nb

cosipy/image_deconvolution/dataIF_COSI_DC2.py

@@ -79,6 +79,8 @@ def load(cls, name, event_binned_data, dict_bkg_binned_data, rsp, coordsys_conv_
         elif isinstance(rsp, Histogram):
             new._image_response = rsp
 
+        new.image_response_contents = new._image_response.contents
+
         # We modify the axes in event, bkg_models, response. This is only for DC2.
         new._modify_axes()
 
@@ -205,7 +207,7 @@ def _calc_exposure_map(self):
 
         if self._coordsys_conv_matrix is None:
             self._exposure_map = Histogram(self._model_axes, unit = self._image_response.unit * u.sr)
-            self._exposure_map[:] = np.sum(self._image_response.contents, axis = (2,3,4)) * self.model_axes['lb'].pixarea()
+            self._exposure_map[:] = np.sum(self.image_response_contents, axis = (2,3,4)) * self.model_axes['lb'].pixarea()
         else:
             self._exposure_map = Histogram(self._model_axes, unit = self._image_response.unit * self._coordsys_conv_matrix.unit * u.sr)
             self._exposure_map[:] = np.tensordot(np.sum(self._coordsys_conv_matrix, axis = (0)), 
@@ -248,7 +250,7 @@ def calc_expectation(self, model, dict_bkg_norm = None, almost_zero = 1e-12):
         expectation = Histogram(self.data_axes)
 
         if self._coordsys_conv_matrix is None:
-            expectation[:] = np.tensordot( model.contents, self._image_response.contents, axes = ([0,1],[0,1])) * model.axes['lb'].pixarea()
+            expectation[:] = np.tensordot( model.contents, self.image_response_contents, axes = ([0,1],[0,1])) * model.axes['lb'].pixarea()
             # ['lb', 'Ei'] x [NuLambda(lb), Ei, Em, Phi, PsiChi] -> [Em, Phi, PsiChi]
         else:
             map_rotated = np.tensordot(self._coordsys_conv_matrix.contents, model.contents, axes = ([1], [0])) 
@@ -257,7 +259,7 @@ def calc_expectation(self, model, dict_bkg_norm = None, almost_zero = 1e-12):
             map_rotated *= model.axes['lb'].pixarea()
             # data.coordsys_conv_matrix.contents is sparse, so the unit should be restored.
             # the unit of map_rotated is 1/cm2 ( = s * 1/cm2/s/sr * sr)
-            expectation[:] = np.tensordot( map_rotated, self._image_response.contents, axes = ([1,2], [0,1]))
+            expectation[:] = np.tensordot( map_rotated, self.image_response_contents, axes = ([1,2], [0,1]))
             # [Time/ScAtt, NuLambda, Ei] x [NuLambda, Ei, Em, Phi, PsiChi] -> [Time/ScAtt, Em, Phi, PsiChi]
 
         if dict_bkg_norm is not None: 
@@ -293,10 +295,10 @@ def calc_T_product(self, dataspace_histogram):
         hist = Histogram(self.model_axes, unit = hist_unit)
 
         if self._coordsys_conv_matrix is None:
-            hist[:] = np.tensordot(dataspace_histogram.contents, self._image_response.contents, axes = ([0,1,2], [2,3,4])) * self.model_axes['lb'].pixarea()
+            hist[:] = np.tensordot(dataspace_histogram.contents, self.image_response_contents, axes = ([0,1,2], [2,3,4])) * self.model_axes['lb'].pixarea()
             # [Em, Phi, PsiChi] x [NuLambda (lb), Ei, Em, Phi, PsiChi] -> [NuLambda (lb), Ei]
         else:
-            _ = np.tensordot(dataspace_histogram.contents, self._image_response.contents, axes = ([1,2,3], [2,3,4])) 
+            _ = np.tensordot(dataspace_histogram.contents, self.image_response_contents, axes = ([1,2,3], [2,3,4])) 
             # [Time/ScAtt, Em, Phi, PsiChi] x [NuLambda, Ei, Em, Phi, PsiChi] -> [Time/ScAtt, NuLambda, Ei]
 
             hist[:] = np.tensordot(self._coordsys_conv_matrix.contents, _, axes = ([0,2], [0,1])) \