Adjusted the batch size to enable llama2-70b to run on v5p-128

[hsuan-lun-chiang]

Description

Currently, the benchmark configuration for Llama2-70b training on v5p-128 encounters OOM issue. This PR reduces the batch size to 2 (previously 4) to prevent OOM errors during training.

Tests

Training with batch size 4
Training with batch size 2

Checklist

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

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• 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, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!