dynamic map checkpoint naming based on model objective.

[xin3he]

Description

This pull request improves the handling of model path remapping during quantization configuration and enhances the evaluation helpers and tests for HuggingFace models, including multimodal support. The main changes ensure remapping is only applied when appropriate, preventing incorrect path transformations for composite models, and extend the evaluation utilities and tests for better flexibility and coverage.

Quantization remapping logic improvements:

• Added logic in get_layer_config (convert_model.py) to only apply checkpoint conversion mapping when the model's module paths do not already match the checkpoint namespace, preventing incorrect remapping for composite models such as VLMs.
• Updated the remapping of quant_block_list and extra_config to respect the new _should_remap flag, ensuring remapping only occurs when necessary.

Evaluation helper enhancements:

• Modified evaluate_accuracy in helpers.py to accept a model_type parameter, allowing evaluation of different model types (e.g., "hf", "hf-multimodal"). Also, now passes the limit parameter to simple_evaluate for more controlled test runs.
  Test improvements:

• Refactored HuggingFace integration tests to use the new model_type and limit parameters, and added a new test for HuggingFace multimodal models (test_hf_multimodal).

Type of Change

Bug fix

Related Issues

Fixes or relates to #1784 #1836

Checklist Before Submitting

• My code has been tested locally.
• Documentation has been updated as needed.
• New or updated tests are included where applicable.
• The CUDA CI has passed. You can trigger it by commenting /azp run Unit-Test-CUDA-AutoRound.

[xin3he]

/azp run Unit-Test-CUDA-AutoRound

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.