check dataset size before training

mindspore/train/model.py

@@ -555,10 +555,11 @@ class Model:
         """
         dataset_sink_mode = Validator.check_bool(dataset_sink_mode)
         Validator.check_is_int(sink_size)
+        dataset_size = train_dataset.get_dataset_size()
+        if dataset_size == 0:
+            raise ValueError("There is no valid data in dataset, please check dataset file first.")
         if sink_size == -1:
-            sink_size = train_dataset.get_dataset_size()
-            if sink_size == 0:
-                raise ValueError("There is no valid data in dataset, please check dataset file first.")
+            sink_size = dataset_size
         if sink_size < -1 or sink_size == 0:
             raise ValueError("The sink_size must be -1 or positive, but got sink_size {}.".format(sink_size))