Fix a bug, use _ceil_to_nearest instead as _round_to_nearest is not d…

[BestJuly]

What does this PR do ?

Fix a bug when training SFT model with dialog dataset when setting self.pad_to_max_length=False

Collection: [Note which collection this PR will affect]

Changelog

#gpt_sft_chat_dataset.py

## original code
## max_length = min(self.max_seq_length, self._round_to_nearest(max_length, 8))

## Changed to
max_length = min(self.max_seq_length, self._ceil_to_nearest(max_length, 8))

In gpt_sft_chat_dataset.py, we use _ceil_to_nearest which is also defined in this file. However, _round_to_nearest is not defined in NeMo.

Usage

• You can potentially add a usage example below

# Add a code snippet demonstrating how to use this 

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
  • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

• New Feature
• Bugfix
• Documentation

If you haven't finished some of the above items you can still open "Draft" PR.