Fix HFInferenceParams missing is_compileable for transformers 5.x compatibility

[svc-bionemo]

Problem

Nightly CI has been failing for 7 consecutive days (4/11–4/17) across models/mixtral, models/llama3, models/qwen, and verify-recipe-tests.

All failures share the same root cause:

AttributeError: 'HFInferenceParams' object has no attribute 'is_compileable'

The 26.03 PyTorch container update (#1540) brought transformers==5.5.4, which added an is_compileable property check on cache objects during generate(). The custom HFInferenceParams class does not implement this property.

Fix

Added is_compileable property returning False to HFInferenceParams in all 6 affected files (4 model sources + 2 recipe copies). The property returns False since this custom TE-based cache wrapper is not compatible with torch.compile.

Failing CI run

https://github.com/NVIDIA/bionemo-framework/actions/runs/24558420210

Testing

• Pre-commit checks pass
• check_copied_files.py confirms all copies are in sync
• Fix is minimal and targeted (5 lines per file)

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 9007c9ff-b348-43b0-a214-d80790d167b6

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[pstjohn]

Nightly CI has been failing for 7 consecutive days (4/11–4/17) across models/mixtral, models/llama3, models/qwen, and verify-recipe-tests.

that's your only job! 😠
bad 🦞