[Data] Allow automatic handling of string features as byte features during TFRecord serialization

[EdwardCuiPeacock]

[Sorry, previous PR was kind of messed up...]

Why are these changes needed?

I am attempting to use ray.data.write_tfrecords to convert my features into TFRecords. Since some of my features are string features, the following error is thrown:

ValueError: Value is of type string, which we cannot convert to a supported tf.train.Feature storage type (bytes, float, or int).

Naturally, TFRecords serializes string features as bytes feature automatically (as typically done in Tensorflow and Tensorflow-extended (TFX); see for example: https://github.com/tensorflow/tfx/blob/fd070288ffa5dcb28204810d2bed261c7e1df201/tfx/components/example_gen/csv_example_gen/executor.py#L67). Making the TFRecord writer automatically treat strings as bytes would help serialize string features correctly, avoiding the hassle of creating additional map functions to convert string features into bytes.

When reading TFRecords, however, we should still keep the bytes features as bytes, as the program does not necessarily know a byte object represents a string. In this example provided by Tensorflow, tensors are serialized into bytes and stored in tfrecords, using tf.io.serialize_tensor.

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[amogkam]

Thanks for the contribution @EdwardCuiPeacock! Do you think you can add a small test for this in test_tfrecords.py? We can add a case for string in test_read_tfrecords

Also cc @scottjlee for review

[EdwardCuiPeacock]

Hi, @amogkam . I added the new unit tests. To install the test_requirement.txt though, I had to change numba from version 0.56.4 to 0.57.1 temporarily, since 0.56.4 does not have a Python 3.11 wheel built (maybe we should upgrade that at some point). But all tests in test_tfrecords.py should be passing now.

[scottjlee]

Thanks for adding this feature!

[amogkam]

Thanks for the contribution @EdwardCuiPeacock! It's been merged.