[NPU] fix VoxtralRealtimeTextModel miss in 5.3.0 transformers version for model Mistral-Small-3.1-2506

[ZeyuanChen2000]

Motivation

In the 5.3.0 transformers version, running model case of test_npu_mistral_small_3_1_24b_instruct_2503.py will report error Could not find VoxtralRealtimeTextModel when call function _config_cls_name_to_arch_name_mapping()

Modifications

In 5.3.0 transformers, _model_mapping becomes _LazyAutoMapping object, use raw_config_mapping and raw_model_mapping to reconstruct the mapping.

Accuracy Tests

Speed Tests and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
• Follow the SGLang code style guidance.

Review and Merge Process

1. Ping Merge Oncalls to start the process. See the PR Merge Process.
2. Get approvals from CODEOWNERS and other reviewers.
3. Trigger CI tests with comments or contact authorized users to do so.
   • Common commands include /tag-and-rerun-ci, /tag-run-ci-label, /rerun-failed-ci
4. After green CI and required approvals, ask Merge Oncalls or people with Write permission to merge the PR.

[gemini-code-assist]

Code Review

This pull request updates the model architecture mapping in llava.py to handle newer transformers versions by accessing internal mapping attributes and adds specific execution arguments to the Mistral model tests on NPU. Feedback suggests documenting the reliance on private library attributes and refactoring the mapping logic into a dictionary comprehension for improved clarity.

[sglang-npu-bot]

/tag-and-rerun-ci

[sglang-npu-bot]

/rerun-failed-ci

[sglang-npu-bot]

/rerun-failed-ci