optimize serving_score loops.

[zhanluxianshen]

Purpose

Test Plan

Test Result

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[gemini-code-assist]

Code Review

The pull request optimizes two loops by using list slicing, which is a good improvement for conciseness and performance. However, this change introduced a critical issue by removing a None check on the embeddings list. This could lead to runtime errors. I've added a review comment with a suggested fix to re-introduce the check in a Pythonic way.

[gemini-code-assist]

This optimization is a good idea, but it removes the crucial assert ... is not None check. The embeddings list is of type list[Optional[PoolingRequestOutput]] and can contain None if an embedding generation fails. Removing the check can lead to a runtime AttributeError in _cosine_similarity when accessing emb.outputs.data.

I suggest re-introducing the assertion in a more Pythonic way that keeps the slicing optimization.

Note that static type checkers might still complain about the types of emb_texts_1 and emb_texts_2 after this change. You may need to add a # type: ignore or use typing.cast if you are using a strict type checking setup.

Suggested change

	emb_texts_1 = embeddings[:len(texts_1)]
	emb_texts_2 = embeddings[len(texts_1):]
	emb_texts_1, emb_texts_2 = embeddings[:len(texts_1)], embeddings[len(texts_1):]
	assert all(e is not None for e in emb_texts_1) and all(e is not None for e in emb_texts_2), "Embedding generation failed, resulting in None values."