Fix image_classification bug

fluid/image_classification/reader.py

@@ -140,7 +140,7 @@ def _reader_creator(file_list,
                 # distributed mode if the env var `PADDLE_TRAINING_ROLE` exits
                 trainer_id = int(os.getenv("PADDLE_TRAINER_ID", "0"))
                 trainer_count = int(os.getenv("PADDLE_TRAINERS", "1"))
-                per_node_lines = len(full_lines) / trainer_count
+                per_node_lines = len(full_lines) // trainer_count
                 lines = full_lines[trainer_id * per_node_lines:(trainer_id + 1)
                                    * per_node_lines]
                 print(