'open_images_v4' with 'array_record' raise ValueError

[seongmoon729]

/!\ PLEASE INCLUDE THE FULL STACKTRACE AND CODE SNIPPET

Short description
When I try to load the 'open_images_v4' dataset with 'array_record' file format, it raises ValueError.

Environment information

• Operating System: ubuntu 20.04

• Python version: 3.8.13

• tensorflow-datasets/tfds-nightly version: tensorflow-datasets=4.9.2

• tensorflow/tf-nightly version: no tensorflow

• Does the issue still exists with the last tfds-nightly package (pip install --upgrade tfds-nightly) ? Yes, I checked with tfds-nightly-4.9.2.dev202308090034.

Reproduction instructions

import os
os.environ.pop('TFDS_DATA_DIR', None)
import tensorflow_datasets as tfds

builder = tfds.builder('open_images_v4', file_format='array_record', data_dir='./tensorflow_datasets/')
builder.download_and_prepare()
ds = builder.as_data_source()['train']
print(ds[0])

Link to logs
Traceback (most recent call last):
File "test_tfds.py", line 27, in
batch = next(ds)
File "/home/moon/.local/miniforge/envs/filtering/lib/python3.8/site-packages/tensorflow_datasets/core/data_sources/array_record.py", line 83, in iter
yield self[i]
File "/home/moon/.local/miniforge/envs/filtering/lib/python3.8/site-packages/tensorflow_datasets/core/data_sources/array_record.py", line 95, in getitem
return self.dataset_info.features.deserialize_example_np(
File "/home/moon/.local/miniforge/envs/filtering/lib/python3.8/site-packages/tensorflow_datasets/core/features/top_level_feature.py", line 159, in deserialize_example_np
return self.decode_example_np(example_data, decoders=decoders)
File "/home/moon/.local/miniforge/envs/filtering/lib/python3.8/site-packages/tensorflow_datasets/core/features/top_level_feature.py", line 112, in decode_example_np
return self._decode_example_generic(
File "/home/moon/.local/miniforge/envs/filtering/lib/python3.8/site-packages/tensorflow_datasets/core/features/top_level_feature.py", line 81, in _decode_example_generic
flatten_decoded = [
File "/home/moon/.local/miniforge/envs/filtering/lib/python3.8/site-packages/tensorflow_datasets/core/features/top_level_feature.py", line 82, in
_decode_feature( # pylint: disable=g-complex-comprehension
File "/home/moon/.local/miniforge/envs/filtering/lib/python3.8/site-packages/tensorflow_datasets/core/features/top_level_feature.py", line 209, in _decode_feature
return decoder.decode_example_np(example, **decode_kwargs)
File "/home/moon/.local/miniforge/envs/filtering/lib/python3.8/site-packages/tensorflow_datasets/core/features/tensor_feature.py", line 234, in decode_example_np
return example_data.reshape(shape)
ValueError: cannot reshape array of size 8 into shape (4,)

Expected behavior
Load the 'open_images_v4' dataset without exception and print a data

Additional context
Add any other context about the problem here.