Remove batch truncation for fractional batch sizes.

[chaserogo]

Description

When running with a per_device_batch_size < 1, rather than splitting larger batches from the data loader into the smaller batch sizes for training, the larger batch is truncated. This PR changes it so that instead the larger batch is split into smaller batches which are looped over instead, so that no data is discarded.

Tests

Please describe how you tested this change, and include any instructions and/or
commands to reproduce.

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.