fix: accept Hugging Face configs without a model_type in HfModel

[Yuri Khrustalev (ykhrustalev)]

Describe your changes

Problem
LiquidAI/LFM2.5-Audio-1.5B has no model_type in its config.json, so AutoConfig rejects it and HfModelHandler can't be constructed. onnxruntime-genai's model builder already supports this checkpoint, so only Olive blocks ModelBuilder.

Solution

• Falls back to a plain PretrainedConfig in get_model_config when the config has no model_type
• Applies load kwargs to the fallback config the same way AutoConfig does
• Keeps raising for unknown model types and for remote-code configs loaded without trust_remote_code

Testing

• Ran ModelBuilder on a tiny random checkpoint with the LFM2.5-Audio config layout: it produced model.onnx and a genai_config.json of type lfm2_audio_text (onnxruntime-genai 0.17.0)

Checklist before requesting a review

• Add unit tests for this change.
• Make sure all tests can pass.
• Update documents if necessary.
• Lint and apply fixes to your code by running lintrunner -a
• Is this a user-facing change? If yes, give a description of this change to be included in the release notes.

Release note: HfModel accepts Hugging Face checkpoints whose config.json has no model_type, such as LiquidAI/LFM2.5-Audio-1.5B, so ModelBuilder can convert them.

[azure-pipelines]

Azure Pipelines:
There may be pipelines that require an authorized user to comment /azp run to run.

[Copilot]

Copilot review overview

🟢 Approval recommended

No unresolved issues were identified, and all reviewed changes are covered by tests.

Review effort: Lite
Findings: None

What changed in this PR

Adds support for Hugging Face checkpoints whose configurations lack model_type, enabling HfModel and ModelBuilder processing.

Changes:

• Falls back to PretrainedConfig while preserving load kwargs.
• Retains validation for unknown types and untrusted remote code.
• Adds focused unit tests.

File	Description
test/​common/​test_hf.py	Tests fallback behavior and error cases.
olive/​common/​hf/​utils.py	Implements configuration fallback logic.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.