Task-specific pipeline init args (#28439)

* Abstract out pipeline init args

* Address PR comments

* Reword

* BC PIPELINE_INIT_ARGS

* Remove old arguments

* Small fix

src/transformers/pipelines/visual_question_answering.py

 from typing import Union
 
 from ..utils import add_end_docstrings, is_torch_available, is_vision_available, logging
-from .base import PIPELINE_INIT_ARGS, Pipeline
+from .base import Pipeline, build_pipeline_init_args
 
 
 if is_vision_available():
@@ -15,7 +15,7 @@ if is_torch_available():
 logger = logging.get_logger(__name__)
 
 
-@add_end_docstrings(PIPELINE_INIT_ARGS)
+@add_end_docstrings(build_pipeline_init_args(has_tokenizer=True, has_image_processor=True))
 class VisualQuestionAnsweringPipeline(Pipeline):
     """
     Visual Question Answering pipeline using a `AutoModelForVisualQuestionAnswering`. This pipeline is currently only

src/transformers/pipelines/zero_shot_audio_classification.py

@@ -23,13 +23,13 @@ from ..utils import (
     logging,
 )
 from .audio_classification import ffmpeg_read
-from .base import PIPELINE_INIT_ARGS, Pipeline
+from .base import Pipeline, build_pipeline_init_args
 
 
 logger = logging.get_logger(__name__)
 
 
-@add_end_docstrings(PIPELINE_INIT_ARGS)
+@add_end_docstrings(build_pipeline_init_args(has_feature_extractor=True, has_tokenizer=True))
 class ZeroShotAudioClassificationPipeline(Pipeline):
     """
     Zero shot audio classification pipeline using `ClapModel`. This pipeline predicts the class of an audio when you

src/transformers/pipelines/zero_shot_classification.py

@@ -5,7 +5,7 @@ import numpy as np
 
 from ..tokenization_utils import TruncationStrategy
 from ..utils import add_end_docstrings, logging
-from .base import PIPELINE_INIT_ARGS, ArgumentHandler, ChunkPipeline
+from .base import ArgumentHandler, ChunkPipeline, build_pipeline_init_args
 
 
 logger = logging.get_logger(__name__)
@@ -43,7 +43,7 @@ class ZeroShotClassificationArgumentHandler(ArgumentHandler):
         return sequence_pairs, sequences
 
 
-@add_end_docstrings(PIPELINE_INIT_ARGS)
+@add_end_docstrings(build_pipeline_init_args(has_tokenizer=True))
 class ZeroShotClassificationPipeline(ChunkPipeline):
     """
     NLI-based zero-shot classification pipeline using a `ModelForSequenceClassification` trained on NLI (natural

src/transformers/pipelines/zero_shot_image_classification.py

@@ -9,7 +9,7 @@ from ..utils import (
     logging,
     requires_backends,
 )
-from .base import PIPELINE_INIT_ARGS, Pipeline
+from .base import Pipeline, build_pipeline_init_args
 
 
 if is_vision_available():
@@ -29,7 +29,7 @@ if is_tf_available():
 logger = logging.get_logger(__name__)
 
 
-@add_end_docstrings(PIPELINE_INIT_ARGS)
+@add_end_docstrings(build_pipeline_init_args(has_image_processor=True))
 class ZeroShotImageClassificationPipeline(Pipeline):
     """
     Zero shot image classification pipeline using `CLIPModel`. This pipeline predicts the class of an image when you

src/transformers/pipelines/zero_shot_object_detection.py

 from typing import Any, Dict, List, Union
 
 from ..utils import add_end_docstrings, is_torch_available, is_vision_available, logging, requires_backends
-from .base import PIPELINE_INIT_ARGS, ChunkPipeline
+from .base import ChunkPipeline, build_pipeline_init_args
 
 
 if is_vision_available():
@@ -19,7 +19,7 @@ if is_torch_available():
 logger = logging.get_logger(__name__)
 
 
-@add_end_docstrings(PIPELINE_INIT_ARGS)
+@add_end_docstrings(build_pipeline_init_args(has_image_processor=True))
 class ZeroShotObjectDetectionPipeline(ChunkPipeline):
     """
     Zero shot object detection pipeline using `OwlViTForObjectDetection`. This pipeline predicts bounding boxes of