Fix minor issues in backend ops

keras/src/backend/jax/nn.py

@@ -273,9 +273,11 @@ def conv(
             f"kernel in_channels {kernel_in_channels}. "
         )
     feature_group_count = channels // kernel_in_channels
+    kernel = convert_to_tensor(kernel)
+    inputs = convert_to_tensor(inputs, dtype=kernel.dtype)
     return jax.lax.conv_general_dilated(
-        convert_to_tensor(inputs),
-        convert_to_tensor(kernel),
+        inputs,
+        kernel,
         strides,
         padding,
         rhs_dilation=dilation_rate,

keras/src/backend/torch/numpy.py

@@ -1251,8 +1251,8 @@ def split(x, indices_or_sections, axis=0):
         dim=axis,
     )
     if dim == 0 and isinstance(indices_or_sections, int):
-        out = tuple(out[0].clone() for _ in range(indices_or_sections))
-    return out
+        out = [out[0].clone() for _ in range(indices_or_sections)]
+    return list(out)
 
 
 def stack(x, axis=0):

keras/src/ops/image.py

@@ -1076,15 +1076,26 @@ def _crop_images(
             f"Received: target_width={target_width}"
         )
 
+    if isinstance(top_cropping, int) and isinstance(left_cropping, int):
+        start_indices = [0, top_cropping, left_cropping, 0]
+    else:
+        start_indices = backend.numpy.stack([0, top_cropping, left_cropping, 0])
+    if (
+        isinstance(batch, int)
+        and isinstance(target_height, int)
+        and isinstance(target_width, int)
+        and isinstance(depth, int)
+    ):
+        shape = [batch, target_height, target_width, depth]
+    else:
+        shape = backend.numpy.stack([batch, target_height, target_width, depth])
     cropped = ops.slice(
         images,
-        backend.numpy.stack([0, top_cropping, left_cropping, 0]),
-        backend.numpy.stack([batch, target_height, target_width, depth]),
+        start_indices,
+        shape,
     )
 
-    cropped_shape = [batch, target_height, target_width, depth]
-    cropped = backend.numpy.reshape(cropped, cropped_shape)
-
+    cropped = backend.numpy.reshape(cropped, shape)
     if not is_batch:
         cropped = backend.numpy.squeeze(cropped, axis=[0])
     return cropped

keras/src/ops/numpy_test.py

@@ -3974,6 +3974,7 @@ def test_sort(self):
 
     def test_split(self):
         x = np.array([[1, 2, 3], [3, 2, 1]])
+        self.assertIsInstance(knp.split(x, 2), list)
         self.assertAllClose(knp.split(x, 2), np.split(x, 2))
         self.assertAllClose(knp.Split(2)(x), np.split(x, 2))
         self.assertAllClose(