Standardize SFT model interface with TunixMaxTextAdapter

[igorts-git]

Description

Overview

This PR refactors train_sft to wrap the model with TunixMaxTextAdapter that is currently used in train_rl to convert model argument names between Tunix and MaxText. This is needed as a preparation work for adding support of DPO.

Tests

• Successfully ran a 2-step SFT job with gemma2-2b to confirm the naming translation works correctly.
• Verified that the model interface remains compatible with existing SFT configurations.

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!