Resolve CPU UT failure with transformers 5.0.0rc

[xin3he]

test/test_cpu got errors as below:

• FAILED test_cpu/test_autoround.py::TestAutoRound::test_not_convert_modules - AttributeError: 'Qwen2VLForConditionalGeneration' object has no attribute 'visual'
• FAILED test_cpu/test_cli_usage.py::TestAutoRoundCmd::test_auto_round_cmd - AssertionError: cmd line test fail, please have a check
• FAILED test_cpu/test_export.py::TestAutoRound::test_static_afp8_export[fp8] - AssertionError: assert torch.bfloat16 == torch.float32
• FAILED test_cpu/test_export.py::TestAutoRound::test_static_fp8_attn - AssertionError: assert torch.bfloat16 == torch.float32
• FAILED test_cpu/test_gguf_format.py::TestGGUF::test_basic_usage - AssertionError: cmd line test fail, please have a check
• FAILED test_cpu/test_gguf_format.py::TestGGUF::test_q4_0 - TypeError: expected str, bytes or os.PathLike object, not NoneType
• FAILED test_cpu/test_gguf_format.py::TestGGUF::test_func - TypeError: expected str, bytes or os.PathLike objec05:55:39 [80/28433]
• FAILED test_cpu/test_gguf_format.py::TestGGUF::test_all_format - AssertionError: cmd line test fail, please have a check
• FAILED test_cpu/test_gguf_format.py::TestGGUF::test_vlm_gguf - KeyError: 'rope_scaling'
• FAILED test_cpu/test_llmc_integration.py::test_oneshot_application[\nquant_stage:\n quant_modifiers:\n AutoRoundModifier:\n ignore: ["lm_head"]\n iters: 1\n config_groups:\n group_0:\n targets:\n - "Linear"\n input_activations: null\n output_activations: null\n weights:\n num_bits: 4\n type: "int"\n symmetric: true\n strategy: group\n group_size: 128\n] - TypeError: LlamaForCausalLM.init() got an unexpected keyword argument 'use_auth_token'
• FAILED test_cpu/test_llmc_integration.py::test_oneshot_application[recipe1] - TypeError: llamaForCausalLM.init() got an unexpected keyword argument 'use_auth_token'
• FAILED test_cpu/test_mllm.py::TestAutoRoundMLLM::test_pure_text_model_check - OSError: Can't load the model for 'facebook/opt-125m'. If you were trying to load it from 'https://huggingface.co/models', m...
• FAILED test_cpu/test_mxfp_nvfp.py::TestAutoRoundFP::test_qwen_moe_quant_infer - NotImplementedError: MoE model with experts type 'Qwen2MoeExperts' is not supported in export.Please file an issue or add su...
• FAILED test_cpu/test_scheme.py::TestAutoRound::test_gguf - AttributeError: 'Qwen3ForCausalLM' object has no attribute 'last_layer_name_to_block_name'
• FAILED test_cpu/test_scheme.py::TestAutoRound::test_mxfp4 - AssertionError: assert 'mx_fp_rceil' == 'mx_fp'
• FAILED test_cpu/test_scheme.py::TestAutoRound::test_all_scheme - AttributeError: 'Qwen3ForCausalLM' object has no attribute 'last_layer_name_to_block_name'
• FAILED test_cpu/test_scheme.py::TestAutoRound::test_scheme_in_layer_config - OSError: Can't load the model for 'facebook/opt-125m'. If you were trying to load it from 'https://huggingface.co/models', m...
• FAILED test_cpu/test_torch_backend.py::TestAutoRoundTorchBackend::test_torch_4bits_asym - RuntimeError: mixed dtype (CPU): expect parameter to have scalar type of Float
• FAILED test_cpu/test_torch_backend.py::TestAutoRoundTorchBackend::test_torch_4bits_sym - OSError: Can't load the model for 'facebook/opt-125m'. If you were trying to load it from 'https://huggingface.co/models', m...
• ERROR test_cpu/test_autoround_export_to_itrex.py::TestAutoroundExport::test_autoround_int_quant - TypeError: GPTJForCausalLM.init() got an unexpected keyword argument 'torchscript'
• ERROR test_cpu/test_autoround_export_to_itrex.py::TestAutoroundExport::test_config - TypeError: GPTJForCausalLM.init() got an unexpected keyword argument 'torchscript'
• ERROR test_cpu/test_autoround_export_to_itrex.py::TestAutoroundExport::test_xpu_export - TypeError: GPTJForCausalLM.init() got an unexpected keyword argument 'torchscript'
• ERROR test_cpu/test_moe_model.py::test_llama4 - OSError: You are trying to access a gated repo.
• ERROR test_cpu/test_mxfp_nvfp.py::TestAutoRoundFP::test_nvfp4_moe_actmax_rtn - ImportError: cannot import name 'is_torch_fx_available' from 'transformers.utils.import_utils' (/home/sdp/xinhe/transformers...
• ERROR test_cpu/test_mxfp_nvfp.py::TestAutoRoundFP::test_nvfp4_moe_actmax_ar - ImportError: cannot import name 'is_torch_fx_available' from 'transformers.utils.import_utils' (/home/sdp/xinhe/transformers...
• ERROR test_cpu/test_mxfp_nvfp.py::TestAutoRoundFP::test_mxfp4_moe_ar - ImportError: cannot import name 'is_torch_fx_available' from 'transformers.utils.import_utils' (/home/sdp/xinhe/transformers...