how to run create coco tf record from code object detection API TF

Question

It's a silly question but I've been struggling with it for way too long now.. I'm using the object detection API from tensorflow.

when I run my script as below I only get outputs names -00000-of-00001 and none as .records.

train_tf_records = _create_tf_record_from_coco_annotations(train_annotations_file, train_image_dir, train_output_path, include_masks=True, num_shards=1)
test_tf_records = _create_tf_record_from_coco_annotations(test_annotations_file, test_image_dir, testdev_output_path, include_masks=True, num_shards=1)
val_tf_records = _create_tf_record_from_coco_annotations(val_annotations_file, val_image_dir, val_output_path, include_masks=True, num_shards=1)

different to when I use the terminal to run it as:

!python /content/models/research/object_detection/dataset_tools/create_coco_tf_record.py \
      --logtostderr \
      --train_image_dir="/content/train/images/" \
      --val_image_dir="/content/valid/images/" \
      --test_image_dir="/content/test/images/" \
      --train_annotations_file="/content/train/annotations/train.json/" \
      --val_annotations_file="/content/valid/annotations/valid.json/" \
      --testdev_annotations_file="/content/test/annotations/test.json/" \
      --output_dir="/content/./"

then I get my outputs as .record

Someone can explain it to me? Thanks a lot!