Add Image To Text Generation pipeline (#18821)

* Add Image2TextGenerationPipeline to supported pipelines

* Add Flax and Tensorflow support

* Add Flax and Tensorflow small tests

* Add default model for Tensorflow

* Add docstring

* Fix doc style

* Add tiny models for pytorch and flax

* Remove flax from pipeline.
Fix tests

* Use ydshieh/vit-gpt2-coco-en as a default for both PyTorch and Tensorflow

* Fix Tensorflow support

Co-authored-by: Olivier Dehaene <olivier@huggingface.co>

docs/source/en/main_classes/pipelines.mdx

@@ -29,6 +29,7 @@ There are two categories of pipeline abstractions to be aware about:
   - [`FillMaskPipeline`]
   - [`ImageClassificationPipeline`]
   - [`ImageSegmentationPipeline`]
+  - [`Image2TextGenerationPipeline`]
   - [`ObjectDetectionPipeline`]
   - [`QuestionAnsweringPipeline`]
   - [`SummarizationPipeline`]
@@ -365,6 +366,12 @@ That should enable you to do all the custom code you want.
     - __call__
     - all
 
+### Image2TextGenerationPipeline
+
+[[autodoc]] Image2TextGenerationPipeline
+    - __call__
+    - all
+
 ### NerPipeline
 
 [[autodoc]] NerPipeline

src/transformers/__init__.py

@@ -384,6 +384,7 @@ _import_structure = {
         "CsvPipelineDataFormat",
         "FeatureExtractionPipeline",
         "FillMaskPipeline",
+        "Image2TextGenerationPipeline",
         "ImageClassificationPipeline",
         "ImageSegmentationPipeline",
         "JsonPipelineDataFormat",
@@ -3191,6 +3192,7 @@ if TYPE_CHECKING:
         CsvPipelineDataFormat,
         FeatureExtractionPipeline,
         FillMaskPipeline,
+        Image2TextGenerationPipeline,
         ImageClassificationPipeline,
         ImageSegmentationPipeline,
         JsonPipelineDataFormat,

src/transformers/pipelines/__init__.py

@@ -53,6 +53,7 @@ from .base import (
 from .conversational import Conversation, ConversationalPipeline
 from .feature_extraction import FeatureExtractionPipeline
 from .fill_mask import FillMaskPipeline
+from .image2text_generation import Image2TextGenerationPipeline
 from .image_classification import ImageClassificationPipeline
 from .image_segmentation import ImageSegmentationPipeline
 from .object_detection import ObjectDetectionPipeline
@@ -90,6 +91,7 @@ if is_tf_available():
         TFAutoModelForSequenceClassification,
         TFAutoModelForTableQuestionAnswering,
         TFAutoModelForTokenClassification,
+        TFAutoModelForVision2Seq,
     )
 
 if is_torch_available():
@@ -118,6 +120,7 @@ if is_torch_available():
         AutoModelForSpeechSeq2Seq,
         AutoModelForTableQuestionAnswering,
         AutoModelForTokenClassification,
+        AutoModelForVision2Seq,
         AutoModelForVisualQuestionAnswering,
     )
 if TYPE_CHECKING:
@@ -302,6 +305,18 @@ SUPPORTED_TASKS = {
         "default": {"model": {"pt": ("facebook/detr-resnet-50-panoptic", "fc15262")}},
         "type": "image",
     },
+    "image2text-generation": {
+        "impl": Image2TextGenerationPipeline,
+        "tf": (TFAutoModelForVision2Seq,) if is_tf_available() else (),
+        "pt": (AutoModelForVision2Seq,) if is_torch_available() else (),
+        "default": {
+            "model": {
+                "pt": ("ydshieh/vit-gpt2-coco-en", "65636df"),
+                "tf": ("ydshieh/vit-gpt2-coco-en", "65636df"),
+            }
+        },
+        "type": "multimodal",
+    },
     "object-detection": {
         "impl": ObjectDetectionPipeline,
         "tf": (),
@@ -317,7 +332,7 @@ NO_TOKENIZER_TASKS = set()
 # any tokenizer/feature_extractor might be use for a given model so we cannot
 # use the statically defined TOKENIZER_MAPPING and FEATURE_EXTRACTOR_MAPPING to
 # see if the model defines such objects or not.
-MULTI_MODEL_CONFIGS = {"VisionTextDualEncoderConfig", "SpeechEncoderDecoderConfig"}
+MULTI_MODEL_CONFIGS = {"SpeechEncoderDecoderConfig", "VisionEncoderDecoderConfig", "VisionTextDualEncoderConfig"}
 for task, values in SUPPORTED_TASKS.items():
     if values["type"] == "text":
         NO_FEATURE_EXTRACTOR_TASKS.add(task)

src/transformers/pipelines/image2text_generation.py

@@ -0,0 +1,96 @@
+from typing import List, Union
+
+from ..utils import (
+    add_end_docstrings,
+    is_tf_available,
+    is_torch_available,
+    is_vision_available,
+    logging,
+    requires_backends,
+)
+from .base import PIPELINE_INIT_ARGS, Pipeline
+
+
+if is_vision_available():
+    from PIL import Image
+
+    from ..image_utils import load_image
+
+if is_tf_available():
+    from ..models.auto.modeling_tf_auto import TF_MODEL_FOR_VISION_2_SEQ_MAPPING
+
+if is_torch_available():
+    from ..models.auto.modeling_auto import MODEL_FOR_VISION_2_SEQ_MAPPING
+
+logger = logging.get_logger(__name__)
+
+
+@add_end_docstrings(PIPELINE_INIT_ARGS)
+class Image2TextGenerationPipeline(Pipeline):
+    """
+    Image2Text Generation pipeline using a `AutoModelForVision2Seq`. This pipeline predicts a caption for a given
+    image.
+
+    This image to text generation pipeline can currently be loaded from pipeline() using the following task identifier:
+    "image2text-generation".
+
+    See the list of available models on
+    [huggingface.co/models](https://huggingface.co/models?pipeline_tag=image-to-text).
+    """
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        requires_backends(self, "vision")
+        self.check_model_type(
+            TF_MODEL_FOR_VISION_2_SEQ_MAPPING if self.framework == "tf" else MODEL_FOR_VISION_2_SEQ_MAPPING
+        )
+
+    def _sanitize_parameters(self, **kwargs):
+        return {}, {}, {}
+
+    def __call__(self, images: Union[str, List[str], "Image.Image", List["Image.Image"]], **kwargs):
+        """
+        Assign labels to the image(s) passed as inputs.
+
+        Args:
+            images (`str`, `List[str]`, `PIL.Image` or `List[PIL.Image]`):
+                The pipeline handles three types of images:
+
+                - A string containing a HTTP(s) link pointing to an image
+                - A string containing a local path to an image
+                - An image loaded in PIL directly
+
+                The pipeline accepts either a single image or a batch of images.
+
+        Return:
+            A list or a list of list of `dict`: Each result comes as a dictionary with the following key:
+
+            - **generated_text** (`str`) -- The generated text.
+        """
+        return super().__call__(images, **kwargs)
+
+    def preprocess(self, image):
+        image = load_image(image)
+        model_inputs = self.feature_extractor(images=image, return_tensors=self.framework)
+        return model_inputs
+
+    def _forward(self, model_inputs):
+        # FIXME: We need to pop here due to a difference in how `generation_utils.py` and `generation_tf_utils.py`
+        #  parse inputs. In the Tensorflow version, `generate` raises an error if we don't use `input_ids` whereas
+        #  the PyTorch version matches it with `self.model.main_input_name` or `self.model.encoder.main_input_name`
+        #  in the `_prepare_model_inputs` method.
+        inputs = model_inputs.pop(self.model.main_input_name)
+        model_outputs = self.model.generate(inputs, **model_inputs)
+        return model_outputs
+
+    def postprocess(self, model_outputs):
+        records = []
+        for output_ids in model_outputs:
+            record = {
+                "generated_text": self.tokenizer.decode(
+                    output_ids,
+                    skip_special_tokens=True,
+                )
+            }
+            records.append(record)
+        return records

tests/pipelines/test_pipelines_image2text_generation.py

@@ -0,0 +1,171 @@
+# Copyright 2022 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+
+from transformers import MODEL_FOR_VISION_2_SEQ_MAPPING, TF_MODEL_FOR_VISION_2_SEQ_MAPPING, is_vision_available
+from transformers.pipelines import pipeline
+from transformers.testing_utils import is_pipeline_test, require_tf, require_torch, require_vision, slow
+
+from .test_pipelines_common import ANY, PipelineTestCaseMeta
+
+
+if is_vision_available():
+    from PIL import Image
+else:
+
+    class Image:
+        @staticmethod
+        def open(*args, **kwargs):
+            pass
+
+
+@is_pipeline_test
+@require_vision
+class Image2TextGenerationPipelineTests(unittest.TestCase, metaclass=PipelineTestCaseMeta):
+    model_mapping = MODEL_FOR_VISION_2_SEQ_MAPPING
+    tf_model_mapping = TF_MODEL_FOR_VISION_2_SEQ_MAPPING
+
+    def get_test_pipeline(self, model, tokenizer, feature_extractor):
+        pipe = pipeline("image2text-generation", model=model, tokenizer=tokenizer, feature_extractor=feature_extractor)
+        examples = [
+            Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png"),
+            "./tests/fixtures/tests_samples/COCO/000000039769.png",
+        ]
+        return pipe, examples
+
+    def run_pipeline_test(self, pipe, examples):
+        outputs = pipe(examples)
+        self.assertEqual(
+            outputs,
+            [
+                [{"generated_text": ANY(str)}],
+                [{"generated_text": ANY(str)}],
+            ],
+        )
+
+    @require_tf
+    def test_small_model_tf(self):
+        pipe = pipeline("image2text-generation", model="hf-internal-testing/tiny-random-vit-gpt2")
+        image = "./tests/fixtures/tests_samples/COCO/000000039769.png"
+
+        outputs = pipe(image)
+        self.assertEqual(
+            outputs,
+            [
+                {
+                    "generated_text": (
+                        " intermedi intermedi intermedi intermedi intermedi "
+                        "explorer explorer explorer explorer explorer explorer "
+                        "explorer medicine medicine medicine medicine medicine "
+                        "medicine medicine"
+                    )
+                },
+            ],
+        )
+
+        outputs = pipe([image, image])
+        self.assertEqual(
+            outputs,
+            [
+                [
+                    {
+                        "generated_text": (
+                            " intermedi intermedi intermedi intermedi intermedi "
+                            "explorer explorer explorer explorer explorer explorer "
+                            "explorer medicine medicine medicine medicine medicine "
+                            "medicine medicine"
+                        )
+                    },
+                ],
+                [
+                    {
+                        "generated_text": (
+                            " intermedi intermedi intermedi intermedi intermedi "
+                            "explorer explorer explorer explorer explorer explorer "
+                            "explorer medicine medicine medicine medicine medicine "
+                            "medicine medicine"
+                        )
+                    },
+                ],
+            ],
+        )
+
+    @require_torch
+    def test_small_model_pt(self):
+        pipe = pipeline("image2text-generation", model="hf-internal-testing/tiny-random-vit-gpt2")
+        image = "./tests/fixtures/tests_samples/COCO/000000039769.png"
+
+        outputs = pipe(image)
+        self.assertEqual(
+            outputs,
+            [
+                {
+                    "generated_text": "growthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthGOGO"
+                },
+            ],
+        )
+
+        outputs = pipe([image, image])
+        self.assertEqual(
+            outputs,
+            [
+                [
+                    {
+                        "generated_text": "growthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthGOGO"
+                    }
+                ],
+                [
+                    {
+                        "generated_text": "growthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthgrowthGOGO"
+                    }
+                ],
+            ],
+        )
+
+    @slow
+    @require_torch
+    def test_large_model_pt(self):
+        pipe = pipeline("image2text-generation", model="ydshieh/vit-gpt2-coco-en")
+        image = "./tests/fixtures/tests_samples/COCO/000000039769.png"
+
+        outputs = pipe(image)
+        self.assertEqual(outputs, [{"generated_text": "a cat laying on a blanket next to a cat laying on a bed "}])
+
+        outputs = pipe([image, image])
+        self.assertEqual(
+            outputs,
+            [
+                [{"generated_text": "a cat laying on a blanket next to a cat laying on a bed "}],
+                [{"generated_text": "a cat laying on a blanket next to a cat laying on a bed "}],
+            ],
+        )
+
+    @slow
+    @require_tf
+    def test_large_model_tf(self):
+        pipe = pipeline("image2text-generation", model="ydshieh/vit-gpt2-coco-en")
+        image = "./tests/fixtures/tests_samples/COCO/000000039769.png"
+
+        outputs = pipe(image)
+        self.assertEqual(outputs, [{"generated_text": "a cat laying on a blanket next to a cat laying on a bed "}])
+
+        outputs = pipe([image, image])
+        self.assertEqual(
+            outputs,
+            [
+                [{"generated_text": "a cat laying on a blanket next to a cat laying on a bed "}],
+                [{"generated_text": "a cat laying on a blanket next to a cat laying on a bed "}],
+            ],
+        )