optimum cant use custom pipelines

[xiaoyao9184]

The config loaded from a HuggingFace model cannot be passed to transformers_pipeline, rendering custom_pipelines unusable. Similarly, externally loading config results in the error optimum.pipelines.pipelines_base.load_ort_pipeline() got multiple values for keyword argument 'config'. Furthermore, since transformers_pipeline does not support the trust_remote_code parameter, proper usage of custom_pipelines remains impossible.

configload in optimum pipeline

optimum/optimum/pipelines/pipelines_base.py

Lines 336 to 339 in 4fdeea7

	config = kwargs.get("config", None)
	if config is None and isinstance(model, str):
	config = AutoConfig.from_pretrained(model, _from_pipeline=task, **hub_kwargs, **kwargs)
	hub_kwargs["_commit_hash"] = config._commit_hash

config also cannot be passed through external kwargs parameters.

optimum/optimum/pipelines/pipelines_base.py

Lines 370 to 383 in 4fdeea7

	model, model_id, tokenizer, feature_extractor = MAPPING_LOADING_FUNC[accelerator](
	model,
	targeted_task,
	load_tokenizer,
	tokenizer,
	feature_extractor,
	load_feature_extractor,
	SUPPORTED_TASKS=supported_tasks,
	config=config,
	hub_kwargs=hub_kwargs,
	token=token,
	*model_kwargs,
	**kwargs,
	)

config not pass to transformers pipeline

optimum/optimum/pipelines/pipelines_base.py

Lines 390 to 397 in 4fdeea7

	return transformers_pipeline(
	task,
	model=model,
	tokenizer=tokenizer,
	feature_extractor=feature_extractor,
	use_fast=use_fast,
	**kwargs,
	)

since the model is already loaded, transformers_pipeline will not load the config.
https://github.com/huggingface/transformers/blob/2e752ead46a8845e8a160d2043c1336447895690/src/transformers/pipelines/__init__.py#L825-L852

    if isinstance(config, str):
        # ...
    elif config is None and isinstance(model, str):
        # ...

As a result, custom_pipelines cannot be used.
https://github.com/huggingface/transformers/blob/2e752ead46a8845e8a160d2043c1336447895690/src/transformers/pipelines/__init__.py#L855-L856

    custom_tasks = {}
    if config is not None and len(getattr(config, "custom_pipelines", {})) > 0:
        custom_tasks = config.custom_pipelines
        # ...

    # Retrieve the task
    if task in custom_tasks:
        normalized_task = task
        targeted_task, task_options = clean_custom_task(custom_tasks[task])
        if pipeline_class is None:
            if not trust_remote_code:
                raise ValueError(
                    "Loading this pipeline requires you to execute the code in the pipeline file in that"
                    " repo on your local machine. Make sure you have read the code there to avoid malicious use, then"
                    " set the option `trust_remote_code=True` to remove this error."
                )
            class_ref = targeted_task["impl"]
            pipeline_class = get_class_from_dynamic_module(
                class_ref,
                model,
                code_revision=code_revision,
                **hub_kwargs,
            )