[Bug][AutoDeploy]: model-coverage entries fail because config rejects use_cache

System Info

• CPU architecture: x86_64
• GPU properties: not captured from CI artifact
• Libraries
  • TensorRT-LLM branch or tag: main (pipeline context)
  • TensorRT-LLM commit: not captured here
  • Container used: AutoDeploy model-coverage CI container
• OS: CI environment
• Additional context: terminal pipeline 52785950, job 327395444

Who can help?

No response

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

Pipeline 52785720 resolved to terminal model-coverage pipeline 52785950.

Affected registry entries in examples/auto_deploy/model_registry/models.yaml:

• apple/OpenELM-270M-Instruct
• apple/OpenELM-1_1B-Instruct
• apple/OpenELM-3B-Instruct

Representative evidence:

1. Open job 327395444.
2. Inspect per-test stdout logs for the three OpenELM tests.
3. Each traceback ends with:

TypeError: OpenELMConfig.post_init() got an unexpected keyword argument 'use_cache'

Relevant stack tail:

File "/opt/tensorrt-llm/tensorrt_llm/_torch/pyexecutor/config_utils.py", line 479, in load_pretrained_config
  model_config = transformers.AutoConfig.from_pretrained(
...
File "/usr/local/lib/python3.12/dist-packages/huggingface_hub/dataclasses.py", line 190, in init
  self.post_init(**additional_kwargs)
TypeError: OpenELMConfig.post_init() got an unexpected keyword argument 'use_cache'

Expected behavior

OpenELM model-coverage entries should load HF config successfully through the current AutoDeploy/llmapi config-loading path.

actual behavior

TensorRT-LLM fails during HF config loading before model execution starts because OpenELMConfig.post_init() rejects the use_cache keyword argument.

additional notes

This reproduces consistently for all three OpenELM instruct entries. The current local registry has those entries disabled with this reason. Likely fix areas are the TensorRT-LLM config-loading path or OpenELM integration path so unsupported kwargs are filtered or normalized before config instantiation.

Before submitting a new issue...

• Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.

Start with tensorrt_llm/_torch/pyexecutor/config_utils.py, especially load_pretrained_config, and inspect the three OpenELM entries in examples/auto_deploy/model_registry/models.yaml. Use pipeline 52785950 and job 327395444 logs to reproduce the config-loading failure. Done means all three OpenELM model-coverage entries load their HF configuration successfully through the current AutoDeploy path.