set default temperature to 0 in generation config

[NathanHB]

No description provided.

[Copilot]

Pull Request Overview

This PR updates the GenerationParameters model to default temperature to 0 (greedy decoding) instead of None.

• Changed temperature from NonNegativeFloat | None = None to NonNegativeFloat = 0
• Removed support for None as a temperature value
• Added inline comment clarifying default greedy decoding behavior

Comments suppressed due to low confidence (3)

src/lighteval/models/model_input.py:39

• Changing the default for temperature from None to 0 is a breaking API change; consider documenting this in the CHANGELOG and bumping the version or adding a deprecation note for the old behavior.

    temperature: NonNegativeFloat = (

src/lighteval/models/model_input.py:39

• Add or update unit tests to cover the new default temperature=0 behavior and verify that greedy decoding is triggered when no custom value is provided.

    temperature: NonNegativeFloat = (

src/lighteval/models/model_input.py:40

• Update the class docstring or external documentation to reflect that temperature now defaults to 0 and that passing None is no longer supported.

        0  # vllm, transformers, tgi, litellm, sglang # if not set, defaults to greedy decoding

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.