python 3 compatible

dmipy/core/modeling_framework.py

@@ -307,7 +307,7 @@ def _prepare_model_properties(self):
 
     def _check_for_double_model_class_instances(self):
         "Checks all models have unique class instances."
-        if len(self.models) != len(np.unique(self.models)):
+        if len(self.models) != len(set(self.models)):
             msg = "Each model in the multi-compartment model must be "
             msg += "instantiated separately. For example, to make a model "
             msg += "with two sticks, the models must be given as "
@@ -641,8 +641,8 @@ def fit(self, acquisition_scheme, data, parameter_initial_guess=None,
             if number_of_processors is None:
                 number_of_processors = cpu_count()
             pool = pp.ProcessPool(number_of_processors)
-            print ('Using parallel processing with {} workers.').format(
-                number_of_processors)
+            print ('Using parallel processing with {} workers.'.format(
+                number_of_processors))
         else:
             fitted_parameters_lin = np.empty(
                 np.r_[N_voxels, N_parameters], dtype=float)
@@ -665,12 +665,12 @@ def fit(self, acquisition_scheme, data, parameter_initial_guess=None,
                 self, self.scheme,
                 parameter_initial_guess, Ns, N_sphere_samples)
             fit_func = Brute2FineOptimizer(self, self.scheme, Ns)
-            print ('Setup brute2fine optimizer in {} seconds').format(
-                time() - start)
+            print ('Setup brute2fine optimizer in {} seconds'.format(
+                time() - start))
         elif solver == 'mix':
             fit_func = MixOptimizer(self, self.scheme, maxiter)
-            print ('Setup MIX optimizer in {} seconds').format(
-                time() - start)
+            print ('Setup MIX optimizer in {} seconds'.format(
+                time() - start))
 
         start = time()
         for idx, pos in enumerate(zip(*mask_pos)):
@@ -690,10 +690,10 @@ def fit(self, acquisition_scheme, data, parameter_initial_guess=None,
                 [p.get() for p in fitted_parameters_lin])
 
         fitting_time = time() - start
-        print ('Fitting of {} voxels complete in {} seconds.').format(
-            len(fitted_parameters_lin), fitting_time)
-        print ('Average of {} seconds per voxel.').format(
-            fitting_time / N_voxels)
+        print ('Fitting of {} voxels complete in {} seconds.'.format(
+            len(fitted_parameters_lin), fitting_time))
+        print ('Average of {} seconds per voxel.'.format(
+            fitting_time / N_voxels))
 
         fitted_parameters = np.zeros_like(x0_, dtype=float)
         fitted_parameters[mask_pos] = (

dmipy/core/tests/test_acquisition_scheme.py

@@ -148,8 +148,8 @@ def test_shell_indices_with_vayring_diffusion_times(Nsamples=10):
     # time are correctly classified in different shells
     bvalues = np.tile(1e9, Nsamples)
     delta = 0.01
-    Delta = np.hstack([np.tile(0.01, len(bvalues) / 2),
-                       np.tile(0.03, len(bvalues) / 2)])
+    Delta = np.hstack([np.tile(0.01, len(bvalues) // 2),
+                       np.tile(0.03, len(bvalues) // 2)])
     gradient_directions = np.tile(np.r_[1., 0., 0.], (Nsamples, 1))
     scheme = acquisition_scheme_from_bvalues(
         bvalues, gradient_directions, delta, Delta)

dmipy/distributions/distribute_models.py

@@ -14,7 +14,7 @@ class DistributedModel:
 
     def _check_for_double_model_class_instances(self):
         "Checks if there are no models with the same class instantiation."
-        if len(self.models) != len(np.unique(self.models)):
+        if len(self.models) != len(set(self.models)):
             msg = "Each model in the multi-compartment model must be "
             msg += "instantiated separately. For example, to make a model "
             msg += "with two sticks, the models must be given as "

dmipy/distributions/distributions.py

@@ -36,7 +36,7 @@
 SPHERE_SPHERICAL = utils.cart2sphere(SPHERE_CARTESIAN)
 log_bingham_normalization_splinefit = np.load(
     join(DATA_PATH,
-         "bingham_normalization_splinefit.npz"))['arr_0']
+         "bingham_normalization_splinefit.npz"), encoding='bytes')['arr_0']
 
 inverse_sh_matrix_kernel = {
     sh_order: np.linalg.pinv(real_sym_sh_mrtrix(

dmipy/optimizers/brute2fine.py

@@ -81,14 +81,14 @@ def precompute_signal_grid(self, model, x0_vector, Ns, N_sphere_samples):
         mu = cart2mu(sphere_vertices)
         grids_per_mu = []
         N_model_fracts = 0
-        parameter_cardinality_items = model.parameter_cardinality.items()
+        parameter_cardinality_items = list(model.parameter_cardinality.items())
         if len(model.models) > 1:
             N_model_fracts = len(model.models)
             parameter_cardinality_items = parameter_cardinality_items[
                 :-N_model_fracts
             ]
 
-        max_cardinality = np.max(model.parameter_cardinality.values())
+        max_cardinality = np.max(list(model.parameter_cardinality.values()))
         for card_counter in range(max_cardinality):
             per_parameter_vectors = []
             counter = 0
@@ -143,7 +143,7 @@ def precompute_signal_grid(self, model, x0_vector, Ns, N_sphere_samples):
                     np.r_[[fract[i] for fract in lin_nested_fractions]])
 
             counter = 0
-            for name, card in model.parameter_cardinality.items()[
+            for name, card in list(model.parameter_cardinality.items())[
                     -N_model_fracts:]:
                 param_dict[name] = lin_fractions[:, counter]
                 counter += 1