Add a compatibility method so callers expecting PreTrainedModel-like …

[sywangyi]

…API don't crash

pytest tests/quantization/compressed_tensors_integration/test_compressed_tensors.py::CompressedTensorsTest

FAILED tests/quantization/compressed_tensors_integration/test_compressed_tensors.py::CompressedTensorsTest::test_llama_8b_fp8 - AttributeError: 'LlamaRotaryEmbedding' object has no attribute 'get_parameter_or_buffer'
FAILED tests/quantization/compressed_tensors_integration/test_compressed_tensors.py::CompressedTensorsTest::test_tinyllama_w4a16 - AttributeError: 'LlamaRotaryEmbedding' object has no attribute 'get_parameter_or_buffer'
FAILED tests/quantization/compressed_tensors_integration/test_compressed_tensors.py::CompressedTensorsTest::test_tinyllama_w8a16 - AttributeError: 'LlamaRotaryEmbedding' object has no attribute 'get_parameter_or_buffer'
FAILED tests/quantization/compressed_tensors_integration/test_compressed_tensors.py::CompressedTensorsTest::test_tinyllama_w8a8 - AttributeError: 'LlamaRotaryEmbedding' object has no attribute 'get_parameter_or_buffer'

call stack

tests/quantization/compressed_tensors_integration/test_compressed_tensors.py:51:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
tests/quantization/compressed_tensors_integration/test_compressed_tensors.py:67: in _test_quantized_model
    quantized_model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto")
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
src/transformers/models/auto/auto_factory.py:373: in from_pretrained
    return model_class.from_pretrained(
src/transformers/modeling_utils.py:278: in _wrapper
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
src/transformers/modeling_utils.py:3967: in from_pretrained
    device_map = _get_device_map(model, device_map, max_memory, hf_quantizer)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
src/transformers/integrations/accelerate.py:407: in _get_device_map
    device_map = infer_auto_device_map(
src/transformers/integrations/accelerate.py:736: in infer_auto_device_map
    current_buffer_size = compute_module_total_buffer_size(module, hf_quantizer)
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
src/transformers/integrations/accelerate.py:260: in compute_module_total_buffer_size
    module_sizes, _ = compute_module_sizes(model, hf_quantizer, buffers_only=True)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
src/transformers/integrations/accelerate.py:240: in compute_module_sizes
    dtype_size = hf_quantizer.param_element_size(model, name)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
src/transformers/quantizers/base.py:182: in param_element_size
    return model.get_parameter_or_buffer(param_name).element_size()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = LlamaRotaryEmbedding(), name = 'get_parameter_or_buffer'

    def __getattr__(self, name: str) -> Union[Tensor, "Module"]:
        if "_parameters" in self.__dict__:
            _parameters = self.__dict__["_parameters"]
            if name in _parameters:
                return _parameters[name]
        if "_buffers" in self.__dict__:
            _buffers = self.__dict__["_buffers"]
            if name in _buffers:
                return _buffers[name]
        if "_modules" in self.__dict__:
            modules = self.__dict__["_modules"]
            if name in modules:
                return modules[name]
>       raise AttributeError(
            f"'{type(self).__name__}' object has no attribute '{name}'"
        )
E       AttributeError: 'LlamaRotaryEmbedding' object has no attribute 'get_parameter_or_buffer'

[sywangyi]

@ydshieh please help review

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: apertus, arcee, aria, bamba, bitnet, blt, chameleon, cohere, cohere2, csm, cwm, dbrx, deepseek_v2, deepseek_v3, dia, diffllama

[Rocketknight1]

Hmmn, I don't think we want to add that method onto every RotaryEmbedding class! This seems like an issue with the compressed_tensors integration maybe, cc @MekkCyber if you have any idea?

[MekkCyber]

Thanks for your PR @sywangyi ! However this will be fixed soon once we merge this PR : https://github.com/huggingface/transformers/pull/42289/files#diff-7f40070336f6d7b1ffe08e654cdf930080e3cbd4dbbcbee2996fabe4ffc1c2b3, it seems like a shorter solution

[sywangyi]

since https://github.com/huggingface/transformers/pull/42289/files#diff-7f40070336f6d7b1ffe08e654cdf930080e3cbd4dbbcbee2996fabe4ffc1c2b3 contains the fix, I will close this PR