bug fix: duplicate padding for offline mode

[uygnef]

Motivation

I identified an issue where input shifting is applied twice in the codebase:

preprocessing.py#L486

train_eagle3.py#L597

This double shifting results in the input_ids and target_ids being offset by 2 positions, causing a mismatch between embeddings and features. This leads to inconsistency between training and inference behavior.

Metric	Before Changes	After Changes	Improvement
Output Throughput	100.57	117.16	+16.46%
Accept Length	1.66	1.93	+16.23%

Modifications

Related Issues

Accuracy Test

Benchmark & Profiling

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://sgl-fru7574.slack.com/archives/C09784E3EN6 to discuss your PR.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!