fix: for tf.function compatibility in tf backend (#28665)

ivy/functional/backends/tensorflow/creation.py

@@ -55,11 +55,11 @@ def arange(
 
     # convert builtin types to tf scalars, as is expected by tf.range
     if isinstance(start, (float, int)):
-        start = tf.constant(start)
+        start = tf.convert_to_tensor(start)
     if isinstance(stop, (float, int)):
-        stop = tf.constant(stop)
+        stop = tf.convert_to_tensor(stop)
     if isinstance(step, (float, int)):
-        step = tf.constant(step)
+        step = tf.convert_to_tensor(step)
 
     if dtype is None:
         if isinstance(start, int) and isinstance(stop, int) and isinstance(step, int):

ivy/functional/backends/tensorflow/experimental/manipulation.py

@@ -398,7 +398,7 @@ def unique_consecutive(
     x_shape = None
     if axis is None:
         x_shape = x.shape
-        x = tf.reshape(x, -1)
+        x = tf.reshape(x, tf.constant([-1]))
         axis = -1
     ndim = len(x.shape)
     if axis < 0:

ivy/functional/frontends/torch/tensor.py

@@ -70,7 +70,7 @@ def dtype(self):
 
     @property
     def shape(self):
-        return Size(self.ivy_array.shape)
+        return Size(ivy.shape(self.ivy_array, as_array=True))
 
     @property
     def real(self):

ivy/functional/ivy/general.py

@@ -2978,7 +2978,10 @@ def _parse_query(query, x_shape, scatter=False):
             [list(query[i].shape) for i in range(0, array_inds[0])]
             + [list(ivy.shape(array_queries[0], as_array=True))]
             + [[] for _ in range(len(array_inds) - 1)]
-            + [list(query[i].shape) for i in range(array_inds[-1] + 1, len(query))]
+            + [
+                list(ivy.shape(query[i], as_array=True))
+                for i in range(array_inds[-1] + 1, len(query))
+            ]
         )
     else:
         target_shape = [list(q.shape) for q in query]