Fix validation and type-checking errors in optimizers

LION/optimizers/EquivariantSolver.py

@@ -46,7 +46,9 @@ def rotation_group(cardinality: int):
         assert 360 % cardinality == 0
         angle_increment = 360 / cardinality
 
-        return [lambda x: TF.rotate(x, i * angle_increment) for i in range(cardinality)]
+        return [
+            lambda x, i=i: TF.rotate(x, i * angle_increment) for i in range(cardinality)
+        ]
 
     @staticmethod
     def default_parameters() -> LIONParameter:

LION/optimizers/GaussianDenoiserSolver.py

@@ -156,7 +156,7 @@ def validate(self):
                     outputs = self.model(y)
                 validation_loss = np.append(
                     validation_loss,
-                    self.validation_fn(y.to(self.device), outputs.to(self.device))
+                    self.validation_fn(outputs.to(self.device), data.to(self.device))
                     .cpu()
                     .numpy(),
                 )

LION/optimizers/LIONsolver.py

@@ -491,7 +491,7 @@ def __check_attribute(
                     warnings.warn(f"Attribute {attr} is not callable")
                     return 2
         # just standrad type chekcking, error or warn, depends of settings
-        elif isinstance(getattr(self, attr), type):
+        elif not isinstance(getattr(self, attr), expected_type):
             if error:
                 raise ValueError(
                     f"Attribute {attr} is not of type {expected_type}, its {type(getattr(self, attr))}"
@@ -599,7 +599,8 @@ def test(self):
                     data = fdk(data, self.op)
                 output = self.model(data.to(self.device))
                 test_loss = np.append(
-                    test_loss, self.testing_fn(output, target.to(self.device))
+                    test_loss,
+                    self.testing_fn(output, target.to(self.device)).cpu().numpy(),
                 )
 
         if self.verbose: