[OpenVINO] Transformers 4.56/4.57 support 2 (awaiting npu testing)

[IlyasMoutawwakil]

What does this PR do?

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[CuriousPanCake]

@IlyasMoutawwakil what are your expectations for merging this PR? We're considering using transformers 4.57 for one of our testing pipelines, so having the corresponding support on optimum-intel's side would be a very good thing.

[IlyasMoutawwakil]

@CuriousPanCake my expectation is that it is good to go 😁
we have seen an occasional / non-consistent segfault with sam model (can't reproduce it locally and rerunning the test passes without a segfault) but it also started happening at the same time as the openvino 2025.4 release, i personally can't find a way to reproduce it locally and the issue happens randomly so, dunno what to do about that.

[rkazants]

there is an issue with gemma models from new transformers reported from GenAI:

Traceback (most recent call last):
  File "/home/jenkins/agent/workspace/DL-Benchmark/prod/WW46-2025.4.0-20398-RC2-Transformers4.57/LO_CPU_acc_wwb_wwb_ref_nat_vs_genai_CPU_ICX/venv/bin/wwb", line 9, in <module>
    sys.exit(main())
  File "/home/jenkins/agent/workspace/DL-Benchmark/prod/WW46-2025.4.0-20398-RC2-Transformers4.57/LO_CPU_acc_wwb_wwb_ref_nat_vs_genai_CPU_ICX/genai/tools/who_what_benchmark/whowhatbench/wwb.py", line 764, in main
    all_metrics_per_question, all_metrics = evaluator.score(
  File "/home/jenkins/agent/workspace/DL-Benchmark/prod/WW46-2025.4.0-20398-RC2-Transformers4.57/LO_CPU_acc_wwb_wwb_ref_nat_vs_genai_CPU_ICX/genai/tools/who_what_benchmark/whowhatbench/visualtext_evaluator.py", line 84, in score
    predictions = self._generate_data(model_or_data, gen_answer_fn, self.generation_config)
  File "/home/jenkins/agent/workspace/DL-Benchmark/prod/WW46-2025.4.0-20398-RC2-Transformers4.57/LO_CPU_acc_wwb_wwb_ref_nat_vs_genai_CPU_ICX/genai/tools/who_what_benchmark/whowhatbench/visualtext_evaluator.py", line 177, in _generate_data
    gen_answer_fn(
  File "/home/jenkins/agent/workspace/DL-Benchmark/prod/WW46-2025.4.0-20398-RC2-Transformers4.57/LO_CPU_acc_wwb_wwb_ref_nat_vs_genai_CPU_ICX/genai/tools/who_what_benchmark/whowhatbench/wwb.py", line 469, in genai_gen_visual_text
    out = model.generate(
RuntimeError: Exception from src/inference/src/cpp/infer_request.cpp:75:
Check '::getPort(port, name, {_impl->get_inputs(), _impl->get_outputs()})' failed at src/inference/src/cpp/infer_request.cpp:77:
Port for tensor name position_ids was not found.

@IlyasMoutawwakil, please double check and clarify why this issue appeared.

Thanks,
Roman

[IlyasMoutawwakil]

@rkazants no idea, there's not enough context in the traceback you shared to know what went wrong exactly