Use TensorType.dtype instead of TensorVariable.dtype in local_elemwise_fusion_op

aesara/tensor/basic_opt.py

@@ -3112,7 +3112,7 @@ def local_elemwise_fusion_op(op_class, max_input_fct=lambda node: 32, maker=None
                         elif ii in tmp_input:
                             tmp_s_input.append(tmp_scalar[tmp_input.index(ii)])
                         else:
-                            tmp = aes.get_scalar_type(ii.dtype).make_variable()
+                            tmp = aes.get_scalar_type(ii.type.dtype).make_variable()
                             try:
                                 tv = get_test_value(ii)
                                 if tv.size > 0:
@@ -3180,7 +3180,7 @@ def local_elemwise_fusion_op(op_class, max_input_fct=lambda node: 32, maker=None
                 if inputs.count(i) == node.inputs.count(i):
                     s = s_inputs[inputs.index(i)]
                 else:
-                    s = aes.get_scalar_type(i.dtype).make_variable()
+                    s = aes.get_scalar_type(i.type.dtype).make_variable()
                     try:
                         if config.compute_test_value != "off":
                             v = get_test_value(i)
@@ -3232,7 +3232,7 @@ your code will run correctly, but may be slower."""
         new_node = maker(node, composite_op)(*inputs).owner
 
         assert len(new_node.outputs) == 1
-        assert node.outputs[0].dtype == new_node.outputs[0].dtype
+        assert node.outputs[0].type.dtype == new_node.outputs[0].type.dtype
 
         if len(new_node.inputs) > max_nb_input:
             _logger.warning(