Add GLPN (#16199)

* First draft

* Fix logits calculation

* Improve tests

* Add copied from statements

* Fix base_model_prefix

* Improve implementation, upload new models

* Update design

* Fix integration test

* Add model to README and toctree

* Add document image

* Apply suggestions from code review

* Apply suggestions from code review
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Add decoder_hidden_size attribute

* Update design of decoder

* Add DepthEstimatorOutput class

* Rename in_index to head_in_index and add feature extractor tests

* Apply suggestions from code review

* Apply suggestions from code review

* Update pretrained model name and add to doc tests

* Remove test.py script

* Update copied from statements and clean up
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

README.md

@@ -265,6 +265,7 @@ Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
 1. **[FlauBERT](https://huggingface.co/docs/transformers/model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
 1. **[FNet](https://huggingface.co/docs/transformers/model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
 1. **[Funnel Transformer](https://huggingface.co/docs/transformers/model_doc/funnel)** (from CMU/Google Brain) released with the paper [Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing](https://arxiv.org/abs/2006.03236) by Zihang Dai, Guokun Lai, Yiming Yang, Quoc V. Le.
+1. **[GLPN](https://huggingface.co/docs/transformers/master/model_doc/glpn)** (from KAIST) released with the paper [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
 1. **[GPT](https://huggingface.co/docs/transformers/model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
 1. **[GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.
 1. **[GPT-J](https://huggingface.co/docs/transformers/model_doc/gptj)** (from EleutherAI) released in the repository [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax/) by Ben Wang and Aran Komatsuzaki.

README_ko.md

@@ -244,6 +244,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[FlauBERT](https://huggingface.co/docs/transformers/model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
 1. **[FNet](https://huggingface.co/docs/transformers/model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
 1. **[Funnel Transformer](https://huggingface.co/docs/transformers/model_doc/funnel)** (from CMU/Google Brain) released with the paper [Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing](https://arxiv.org/abs/2006.03236) by Zihang Dai, Guokun Lai, Yiming Yang, Quoc V. Le.
+1. **[GLPN](https://huggingface.co/docs/transformers/master/model_doc/glpn)** (from KAIST) released with the paper [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
 1. **[GPT](https://huggingface.co/docs/transformers/model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
 1. **[GPT Neo](https://huggingface.co/docs/transformers/model_doc/gpt_neo)** (from EleutherAI) released in the repository [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
 1. **[GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.

README_zh-hans.md

@@ -268,6 +268,7 @@ conda install -c huggingface transformers
 1. **[FlauBERT](https://huggingface.co/docs/transformers/model_doc/flaubert)** (来自 CNRS) 伴随论文 [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) 由 Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab 发布。
 1. **[FNet](https://huggingface.co/docs/transformers/model_doc/fnet)** (来自 Google Research) 伴随论文 [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) 由 James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon 发布。
 1. **[Funnel Transformer](https://huggingface.co/docs/transformers/model_doc/funnel)** (来自 CMU/Google Brain) 伴随论文 [Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing](https://arxiv.org/abs/2006.03236) 由 Zihang Dai, Guokun Lai, Yiming Yang, Quoc V. Le 发布。
+1. **[GLPN](https://huggingface.co/docs/transformers/master/model_doc/glpn)** (来自 KAIST) 伴随论文 [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) 由 Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim 发布。
 1. **[GPT](https://huggingface.co/docs/transformers/model_doc/openai-gpt)** (来自 OpenAI) 伴随论文 [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) 由 Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever 发布。
 1. **[GPT Neo](https://huggingface.co/docs/transformers/model_doc/gpt_neo)** (来自 EleutherAI) 随仓库 [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) 发布。作者为 Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy 发布。
 1. **[GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2)** (来自 OpenAI) 伴随论文 [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) 由 Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever** 发布。

README_zh-hant.md

@@ -280,6 +280,7 @@ conda install -c huggingface transformers
 1. **[FlauBERT](https://huggingface.co/docs/transformers/model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
 1. **[FNet](https://huggingface.co/docs/transformers/model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
 1. **[Funnel Transformer](https://huggingface.co/docs/transformers/model_doc/funnel)** (from CMU/Google Brain) released with the paper [Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing](https://arxiv.org/abs/2006.03236) by Zihang Dai, Guokun Lai, Yiming Yang, Quoc V. Le.
+1. **[GLPN](https://huggingface.co/docs/transformers/master/model_doc/glpn)** (from KAIST) released with the paper [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
 1. **[GPT](https://huggingface.co/docs/transformers/model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
 1. **[GPT Neo](https://huggingface.co/docs/transformers/model_doc/gpt_neo)** (from EleutherAI) released in the repository [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
 1. **[GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.

docs/source/_toctree.yml

@@ -218,6 +218,8 @@
       title: FSMT
     - local: model_doc/funnel
       title: Funnel Transformer
+    - local: model_doc/glpn
+      title: GLPN
     - local: model_doc/herbert
       title: HerBERT
     - local: model_doc/ibert

docs/source/index.mdx

@@ -89,6 +89,7 @@ conversion utilities for the following models.
 1. **[FlauBERT](model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
 1. **[FNet](model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
 1. **[Funnel Transformer](model_doc/funnel)** (from CMU/Google Brain) released with the paper [Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing](https://arxiv.org/abs/2006.03236) by Zihang Dai, Guokun Lai, Yiming Yang, Quoc V. Le.
+1. **[GLPN](model_doc/glpn)** (from KAIST) released with the paper [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
 1. **[GPT](model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
 1. **[GPT-2](model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.
 1. **[GPT-J](model_doc/gptj)** (from EleutherAI) released in the repository [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax/) by Ben Wang and Aran Komatsuzaki.
@@ -200,6 +201,7 @@ Flax), PyTorch, and/or TensorFlow.
 |          FlauBERT           |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
 |            FNet             |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
 |     Funnel Transformer      |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|            GLPN             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |           GPT Neo           |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
 |            GPT-J            |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
 |           Hubert            |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |

docs/source/model_doc/glpn.mdx

+<!--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.
+-->
+
+# GLPN
+
+<Tip>
+
+This is a recently introduced model so the API hasn't been tested extensively. There may be some bugs or slight
+breaking changes to fix it in the future. If you see something strange, file a [Github Issue](https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title).
+
+</Tip>
+
+## Overview
+
+The GLPN model was proposed in [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436)  by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
+GLPN combines [SegFormer](segformer)'s hierarchical mix-Transformer with a lightweight decoder for monocular depth estimation. The proposed decoder shows better performance than the previously proposed decoders, with considerably
+less computational complexity.
+
+The abstract from the paper is the following:
+
+*Depth estimation from a single image is an important task that can be applied to various fields in computer vision, and has grown rapidly with the development of convolutional neural networks. In this paper, we propose a novel structure and training strategy for monocular depth estimation to further improve the prediction accuracy of the network. We deploy a hierarchical transformer encoder to capture and convey the global context, and design a lightweight yet powerful decoder to generate an estimated depth map while considering local connectivity. By constructing connected paths between multi-scale local features and the global decoding stream with our proposed selective feature fusion module, the network can integrate both representations and recover fine details. In addition, the proposed decoder shows better performance than the previously proposed decoders, with considerably less computational complexity. Furthermore, we improve the depth-specific augmentation method by utilizing an important observation in depth estimation to enhance the model. Our network achieves state-of-the-art performance over the challenging depth dataset NYU Depth V2. Extensive experiments have been conducted to validate and show the effectiveness of the proposed approach. Finally, our model shows better generalisation ability and robustness than other comparative models.*
+
+Tips:
+
+- One can use [`GLPNFeatureExtractor`] to prepare images for the model.
+
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/glpn_architecture.jpg"
+alt="drawing" width="600"/> 
+
+<small> Summary of the approach. Taken from the <a href="https://arxiv.org/abs/2201.07436" target="_blank">original paper</a>. </small>
+
+This model was contributed by [niels](<https://huggingface.co/nielsr). The original code can be found [here](https://github.com/vinvino02/GLPDepth).
+
+## GLPNConfig
+
+[[autodoc]] GLPNConfig
+
+## GLPNFeatureExtractor
+
+[[autodoc]] GLPNFeatureExtractor
+    - __call__
+
+## GLPNModel
+
+[[autodoc]] GLPNModel
+    - forward
+
+## GLPNForDepthEstimation
+
+[[autodoc]] GLPNForDepthEstimation
+    - forward
\ No newline at end of file

src/transformers/__init__.py

@@ -224,6 +224,7 @@ _import_structure = {
     "models.fnet": ["FNET_PRETRAINED_CONFIG_ARCHIVE_MAP", "FNetConfig", "FNetTokenizer"],
     "models.fsmt": ["FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP", "FSMTConfig", "FSMTTokenizer"],
     "models.funnel": ["FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP", "FunnelConfig", "FunnelTokenizer"],
+    "models.glpn": ["GLPN_PRETRAINED_CONFIG_ARCHIVE_MAP", "GLPNConfig"],
     "models.gpt2": ["GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP", "GPT2Config", "GPT2Tokenizer"],
     "models.gpt_neo": ["GPT_NEO_PRETRAINED_CONFIG_ARCHIVE_MAP", "GPTNeoConfig"],
     "models.gptj": ["GPTJ_PRETRAINED_CONFIG_ARCHIVE_MAP", "GPTJConfig"],
@@ -525,6 +526,7 @@ if is_vision_available():
     _import_structure["models.convnext"].append("ConvNextFeatureExtractor")
     _import_structure["models.deit"].append("DeiTFeatureExtractor")
     _import_structure["models.detr"].append("DetrFeatureExtractor")
+    _import_structure["models.glpn"].append("GLPNFeatureExtractor")
     _import_structure["models.imagegpt"].append("ImageGPTFeatureExtractor")
     _import_structure["models.layoutlmv2"].append("LayoutLMv2FeatureExtractor")
     _import_structure["models.layoutlmv2"].append("LayoutLMv2Processor")
@@ -993,6 +995,14 @@ if is_torch_available():
             "load_tf_weights_in_funnel",
         ]
     )
+    _import_structure["models.glpn"].extend(
+        [
+            "GLPN_PRETRAINED_MODEL_ARCHIVE_LIST",
+            "GLPNForDepthEstimation",
+            "GLPNModel",
+            "GLPNPreTrainedModel",
+        ]
+    )
     _import_structure["models.gpt2"].extend(
         [
             "GPT2_PRETRAINED_MODEL_ARCHIVE_LIST",
@@ -2550,6 +2560,7 @@ if TYPE_CHECKING:
     from .models.fnet import FNET_PRETRAINED_CONFIG_ARCHIVE_MAP, FNetConfig, FNetTokenizer
     from .models.fsmt import FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP, FSMTConfig, FSMTTokenizer
     from .models.funnel import FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP, FunnelConfig, FunnelTokenizer
+    from .models.glpn import GLPN_PRETRAINED_CONFIG_ARCHIVE_MAP, GLPNConfig
     from .models.gpt2 import GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP, GPT2Config, GPT2Tokenizer
     from .models.gpt_neo import GPT_NEO_PRETRAINED_CONFIG_ARCHIVE_MAP, GPTNeoConfig
     from .models.gptj import GPTJ_PRETRAINED_CONFIG_ARCHIVE_MAP, GPTJConfig
@@ -2803,6 +2814,7 @@ if TYPE_CHECKING:
         from .models.convnext import ConvNextFeatureExtractor
         from .models.deit import DeiTFeatureExtractor
         from .models.detr import DetrFeatureExtractor
+        from .models.glpn import GLPNFeatureExtractor
         from .models.imagegpt import ImageGPTFeatureExtractor
         from .models.layoutlmv2 import LayoutLMv2FeatureExtractor, LayoutLMv2Processor
         from .models.layoutxlm import LayoutXLMProcessor
@@ -2841,6 +2853,7 @@ if TYPE_CHECKING:
         from .utils.dummy_scatter_objects import *
 
     if is_torch_available():
+
         # Benchmarks
         from .benchmark.benchmark import PyTorchBenchmark
         from .benchmark.benchmark_args import PyTorchBenchmarkArguments
@@ -3195,6 +3208,12 @@ if TYPE_CHECKING:
             FunnelPreTrainedModel,
             load_tf_weights_in_funnel,
         )
+        from .models.glpn import (
+            GLPN_PRETRAINED_MODEL_ARCHIVE_LIST,
+            GLPNForDepthEstimation,
+            GLPNModel,
+            GLPNPreTrainedModel,
+        )
         from .models.gpt2 import (
             GPT2_PRETRAINED_MODEL_ARCHIVE_LIST,
             GPT2DoubleHeadsModel,

src/transformers/modeling_outputs.py

@@ -878,3 +878,33 @@ class ImageClassifierOutput(ModelOutput):
     logits: torch.FloatTensor = None
     hidden_states: Optional[Tuple[torch.FloatTensor]] = None
     attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+@dataclass
+class DepthEstimatorOutput(ModelOutput):
+    """
+    Base class for outputs of depth estimation models.
+
+    Args:
+        loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
+            Classification (or regression if config.num_labels==1) loss.
+        predicted_depth (`torch.FloatTensor` of shape `(batch_size, height, width)`):
+            Predicted depth for each pixel.
+
+        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
+            Tuple of `torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer) of
+            shape `(batch_size, num_channels, height, width)`.
+
+            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
+        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, patch_size,
+            sequence_length)`.
+
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
+    """
+
+    loss: Optional[torch.FloatTensor] = None
+    predicted_depth: torch.FloatTensor = None
+    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    attentions: Optional[Tuple[torch.FloatTensor]] = None

src/transformers/models/__init__.py

@@ -55,6 +55,7 @@ from . import (
     fnet,
     fsmt,
     funnel,
+    glpn,
     gpt2,
     gpt_neo,
     gptj,

src/transformers/models/auto/configuration_auto.py

@@ -30,6 +30,7 @@ logger = logging.get_logger(__name__)
 CONFIG_MAPPING_NAMES = OrderedDict(
     [
         # Add configs here
+        ("glpn", "GLPNConfig"),
         ("maskformer", "MaskFormerConfig"),
         ("poolformer", "PoolFormerConfig"),
         ("convnext", "ConvNextConfig"),
@@ -132,6 +133,7 @@ CONFIG_MAPPING_NAMES = OrderedDict(
 CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict(
     [
         # Add archive maps here
+        ("glpn", "GLPN_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("maskformer", "MASKFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("poolformer", "POOLFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("convnext", "CONVNEXT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
@@ -221,6 +223,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict(
 MODEL_NAMES_MAPPING = OrderedDict(
     [
         # Add full (and cased) model names here
+        ("glpn", "GLPN"),
         ("maskformer", "MaskFormer"),
         ("poolformer", "PoolFormer"),
         ("convnext", "ConvNext"),

src/transformers/models/auto/modeling_auto.py

@@ -28,6 +28,7 @@ logger = logging.get_logger(__name__)
 MODEL_MAPPING_NAMES = OrderedDict(
     [
         # Base model mapping
+        ("glpn", "GLPNModel"),
         ("maskformer", "MaskFormerModel"),
         ("poolformer", "PoolFormerModel"),
         ("convnext", "ConvNextModel"),

src/transformers/models/glpn/__init__.py

+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+# 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.
+from typing import TYPE_CHECKING
+
+# rely on isort to merge the imports
+from ...file_utils import _LazyModule, is_torch_available, is_vision_available
+
+
+_import_structure = {
+    "configuration_glpn": ["GLPN_PRETRAINED_CONFIG_ARCHIVE_MAP", "GLPNConfig"],
+}
+
+if is_vision_available():
+    _import_structure["feature_extraction_glpn"] = ["GLPNFeatureExtractor"]
+
+if is_torch_available():
+    _import_structure["modeling_glpn"] = [
+        "GLPN_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "GLPNForDepthEstimation",
+        "GLPNLayer",
+        "GLPNModel",
+        "GLPNPreTrainedModel",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_glpn import GLPN_PRETRAINED_CONFIG_ARCHIVE_MAP, GLPNConfig
+
+    if is_vision_available():
+        from .feature_extraction_glpn import GLPNFeatureExtractor
+
+    if is_torch_available():
+        from .modeling_glpn import (
+            GLPN_PRETRAINED_MODEL_ARCHIVE_LIST,
+            GLPNForDepthEstimation,
+            GLPNLayer,
+            GLPNModel,
+            GLPNPreTrainedModel,
+        )
+
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)

src/transformers/models/glpn/configuration_glpn.py

+# coding=utf-8
+# Copyright 2022 KAIST and The HuggingFace Inc. 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.
+""" GLPN model configuration"""
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+GLPN_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "vinvino02/glpn-kitti": "https://huggingface.co/vinvino02/gdpdepth-kitti/resolve/main/config.json",
+    # See all GLPN models at https://huggingface.co/models?filter=gdpdepth
+}
+
+
+class GLPNConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`GLPNModel`]. It is used to instantiate an GLPN
+    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a similar configuration to that of the GLPN
+    [kaist/gdpdepth-kitti](https://huggingface.co/kaist/gdpdepth-kitti) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        num_channels (`int`, *optional*, defaults to 3):
+            The number of input channels.
+        num_encoder_blocks (`int`, *optional*, defaults to 4):
+            The number of encoder blocks (i.e. stages in the Mix Transformer encoder).
+        depths (`List[int]`, *optional*, defaults to `[2, 2, 2, 2]`):
+            The number of layers in each encoder block.
+        sr_ratios (`List[int]`, *optional*, defaults to `[8, 4, 2, 1]`):
+            Sequence reduction ratios in each encoder block.
+        hidden_sizes (`List[int]`, *optional*, defaults to `[32, 64, 160, 256]`):
+            Dimension of each of the encoder blocks.
+        patch_sizes (`List[int]`, *optional*, defaults to `[7, 3, 3, 3]`):
+            Patch size before each encoder block.
+        strides (`List[int]`, *optional*, defaults to `[4, 2, 2, 2]`):
+            Stride before each encoder block.
+        num_attention_heads (`List[int]`, *optional*, defaults to `[1, 2, 4, 8]`):
+            Number of attention heads for each attention layer in each block of the Transformer encoder.
+        mlp_ratios (`List[int]`, *optional*, defaults to `[4, 4, 4, 4]`):
+            Ratio of the size of the hidden layer compared to the size of the input layer of the Mix FFNs in the
+            encoder blocks.
+        hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"selu"` and `"gelu_new"` are supported.
+        hidden_dropout_prob (`float`, *optional*, defaults to 0.0):
+            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
+        attention_probs_dropout_prob (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        drop_path_rate (`float`, *optional*, defaults to 0.1):
+            The dropout probability for stochastic depth, used in the blocks of the Transformer encoder.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-6):
+            The epsilon used by the layer normalization layers.
+        decoder_hidden_size (`int`, *optional*, defaults to 32):
+            The dimension of the decoder.
+        max_depth (`int`, *optional*, defaults to 10):
+            The maximum depth of the decoder.
+        head_in_index (`int`, *optional*, defaults to -1):
+            The index of the features to use in the head.
+
+    Example:
+
+    ```python
+    >>> from transformers import GLPNModel, GLPNConfig
+
+    >>> # Initializing a GLPN kaist/gdpdepth-kitti style configuration
+    >>> configuration = GLPNConfig()
+
+    >>> # Initializing a model from the kaist/gdpdepth-kitti style configuration
+    >>> model = GLPNModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+    model_type = "glpn"
+
+    def __init__(
+        self,
+        num_channels=3,
+        num_encoder_blocks=4,
+        depths=[2, 2, 2, 2],
+        sr_ratios=[8, 4, 2, 1],
+        hidden_sizes=[32, 64, 160, 256],
+        patch_sizes=[7, 3, 3, 3],
+        strides=[4, 2, 2, 2],
+        num_attention_heads=[1, 2, 5, 8],
+        mlp_ratios=[4, 4, 4, 4],
+        hidden_act="gelu",
+        hidden_dropout_prob=0.0,
+        attention_probs_dropout_prob=0.0,
+        initializer_range=0.02,
+        drop_path_rate=0.1,
+        layer_norm_eps=1e-6,
+        is_encoder_decoder=False,
+        decoder_hidden_size=64,
+        max_depth=10,
+        head_in_index=-1,
+        **kwargs
+    ):
+        super().__init__(**kwargs)
+
+        self.num_channels = num_channels
+        self.num_encoder_blocks = num_encoder_blocks
+        self.depths = depths
+        self.sr_ratios = sr_ratios
+        self.hidden_sizes = hidden_sizes
+        self.patch_sizes = patch_sizes
+        self.strides = strides
+        self.mlp_ratios = mlp_ratios
+        self.num_attention_heads = num_attention_heads
+        self.hidden_act = hidden_act
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.initializer_range = initializer_range
+        self.drop_path_rate = drop_path_rate
+        self.layer_norm_eps = layer_norm_eps
+        self.decoder_hidden_size = decoder_hidden_size
+        self.max_depth = max_depth
+        self.head_in_index = head_in_index

src/transformers/models/glpn/convert_glpn_to_pytorch.py

+# coding=utf-8
+# Copyright 2022 The HuggingFace Inc. team.
+#
+# 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.
+"""Convert GLPN checkpoints."""
+
+
+import argparse
+from collections import OrderedDict
+from pathlib import Path
+
+import torch
+from PIL import Image
+
+import requests
+from transformers import GLPNConfig, GLPNFeatureExtractor, GLPNForDepthEstimation
+from transformers.utils import logging
+
+
+logging.set_verbosity_info()
+logger = logging.get_logger(__name__)
+
+
+def rename_keys(state_dict):
+    new_state_dict = OrderedDict()
+    for key, value in state_dict.items():
+        if key.startswith("module.encoder"):
+            key = key.replace("module.encoder", "glpn.encoder")
+        if key.startswith("module.decoder"):
+            key = key.replace("module.decoder", "decoder.stages")
+        if "patch_embed" in key:
+            # replace for example patch_embed1 by patch_embeddings.0
+            idx = key[key.find("patch_embed") + len("patch_embed")]
+            key = key.replace(f"patch_embed{idx}", f"patch_embeddings.{int(idx)-1}")
+        if "norm" in key:
+            key = key.replace("norm", "layer_norm")
+        if "glpn.encoder.layer_norm" in key:
+            # replace for example layer_norm1 by layer_norm.0
+            idx = key[key.find("glpn.encoder.layer_norm") + len("glpn.encoder.layer_norm")]
+            key = key.replace(f"layer_norm{idx}", f"layer_norm.{int(idx)-1}")
+        if "layer_norm1" in key:
+            key = key.replace("layer_norm1", "layer_norm_1")
+        if "layer_norm2" in key:
+            key = key.replace("layer_norm2", "layer_norm_2")
+        if "block" in key:
+            # replace for example block1 by block.0
+            idx = key[key.find("block") + len("block")]
+            key = key.replace(f"block{idx}", f"block.{int(idx)-1}")
+        if "attn.q" in key:
+            key = key.replace("attn.q", "attention.self.query")
+        if "attn.proj" in key:
+            key = key.replace("attn.proj", "attention.output.dense")
+        if "attn" in key:
+            key = key.replace("attn", "attention.self")
+        if "fc1" in key:
+            key = key.replace("fc1", "dense1")
+        if "fc2" in key:
+            key = key.replace("fc2", "dense2")
+        if "linear_pred" in key:
+            key = key.replace("linear_pred", "classifier")
+        if "linear_fuse" in key:
+            key = key.replace("linear_fuse.conv", "linear_fuse")
+            key = key.replace("linear_fuse.bn", "batch_norm")
+        if "linear_c" in key:
+            # replace for example linear_c4 by linear_c.3
+            idx = key[key.find("linear_c") + len("linear_c")]
+            key = key.replace(f"linear_c{idx}", f"linear_c.{int(idx)-1}")
+        if "bot_conv" in key:
+            key = key.replace("bot_conv", "0.convolution")
+        if "skip_conv1" in key:
+            key = key.replace("skip_conv1", "1.convolution")
+        if "skip_conv2" in key:
+            key = key.replace("skip_conv2", "2.convolution")
+        if "fusion1" in key:
+            key = key.replace("fusion1", "1.fusion")
+        if "fusion2" in key:
+            key = key.replace("fusion2", "2.fusion")
+        if "fusion3" in key:
+            key = key.replace("fusion3", "3.fusion")
+        if "fusion" in key and "conv" in key:
+            key = key.replace("conv", "convolutional_layer")
+        if key.startswith("module.last_layer_depth"):
+            key = key.replace("module.last_layer_depth", "head.head")
+        new_state_dict[key] = value
+
+    return new_state_dict
+
+
+def read_in_k_v(state_dict, config):
+    # for each of the encoder blocks:
+    for i in range(config.num_encoder_blocks):
+        for j in range(config.depths[i]):
+            # read in weights + bias of keys and values (which is a single matrix in the original implementation)
+            kv_weight = state_dict.pop(f"glpn.encoder.block.{i}.{j}.attention.self.kv.weight")
+            kv_bias = state_dict.pop(f"glpn.encoder.block.{i}.{j}.attention.self.kv.bias")
+            # next, add keys and values (in that order) to the state dict
+            state_dict[f"glpn.encoder.block.{i}.{j}.attention.self.key.weight"] = kv_weight[
+                : config.hidden_sizes[i], :
+            ]
+            state_dict[f"glpn.encoder.block.{i}.{j}.attention.self.key.bias"] = kv_bias[: config.hidden_sizes[i]]
+            state_dict[f"glpn.encoder.block.{i}.{j}.attention.self.value.weight"] = kv_weight[
+                config.hidden_sizes[i] :, :
+            ]
+            state_dict[f"glpn.encoder.block.{i}.{j}.attention.self.value.bias"] = kv_bias[config.hidden_sizes[i] :]
+
+
+# We will verify our results on a COCO image
+def prepare_img():
+    url = "http://images.cocodataset.org/val2017/000000039769.jpg"
+    image = Image.open(requests.get(url, stream=True).raw)
+
+    return image
+
+
+@torch.no_grad()
+def convert_glpn_checkpoint(checkpoint_path, pytorch_dump_folder_path, push_to_hub=False, model_name=None):
+    """
+    Copy/paste/tweak model's weights to our GLPN structure.
+    """
+
+    # load GLPN configuration (Segformer-B4 size)
+    config = GLPNConfig(hidden_sizes=[64, 128, 320, 512], decoder_hidden_size=64, depths=[3, 8, 27, 3])
+
+    # load feature extractor (only resize + rescale)
+    feature_extractor = GLPNFeatureExtractor()
+
+    # prepare image
+    image = prepare_img()
+    pixel_values = feature_extractor(images=image, return_tensors="pt").pixel_values
+
+    logger.info("Converting model...")
+
+    # load original state dict
+    state_dict = torch.load(checkpoint_path, map_location=torch.device("cpu"))
+
+    # rename keys
+    state_dict = rename_keys(state_dict)
+
+    # key and value matrices need special treatment
+    read_in_k_v(state_dict, config)
+
+    # create HuggingFace model and load state dict
+    model = GLPNForDepthEstimation(config)
+    model.load_state_dict(state_dict)
+    model.eval()
+
+    # forward pass
+    outputs = model(pixel_values)
+    predicted_depth = outputs.predicted_depth
+
+    # verify output
+    if model_name is not None:
+        if "nyu" in model_name:
+            expected_slice = torch.tensor(
+                [[4.4147, 4.0873, 4.0673], [3.7890, 3.2881, 3.1525], [3.7674, 3.5423, 3.4913]]
+            )
+        elif "kitti" in model_name:
+            expected_slice = torch.tensor(
+                [[3.4291, 2.7865, 2.5151], [3.2841, 2.7021, 2.3502], [3.1147, 2.4625, 2.2481]]
+            )
+        else:
+            raise ValueError(f"Unknown model name: {model_name}")
+
+        expected_shape = torch.Size([1, 480, 640])
+
+        assert predicted_depth.shape == expected_shape
+        assert torch.allclose(predicted_depth[0, :3, :3], expected_slice, atol=1e-4)
+        print("Looks ok!")
+
+    # finally, push to hub if required
+    if push_to_hub:
+        logger.info("Pushing model and feature extractor to the hub...")
+        model.push_to_hub(
+            repo_path_or_name=Path(pytorch_dump_folder_path, model_name),
+            organization="nielsr",
+            commit_message="Add model",
+            use_temp_dir=True,
+        )
+        feature_extractor.push_to_hub(
+            repo_path_or_name=Path(pytorch_dump_folder_path, model_name),
+            organization="nielsr",
+            commit_message="Add feature extractor",
+            use_temp_dir=True,
+        )
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--checkpoint_path",
+        default=None,
+        type=str,
+        help="Path to the original PyTorch checkpoint (.pth file).",
+    )
+    parser.add_argument(
+        "--pytorch_dump_folder_path", default=None, type=str, help="Path to the folder to output PyTorch model."
+    )
+    parser.add_argument(
+        "--push_to_hub", action="store_true", help="Whether to upload the model to the HuggingFace hub."
+    )
+    parser.add_argument(
+        "--model_name",
+        default="glpn-kitti",
+        type=str,
+        help="Name of the model in case you're pushing to the hub.",
+    )
+    args = parser.parse_args()
+    convert_glpn_checkpoint(args.checkpoint_path, args.pytorch_dump_folder_path, args.push_to_hub, args.model_name)

src/transformers/models/glpn/feature_extraction_glpn.py

+# coding=utf-8
+# Copyright 2022 The HuggingFace Inc. 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.
+"""Feature extractor class for GLPN."""
+
+from typing import Optional, Union
+
+import numpy as np
+from PIL import Image
+
+from ...feature_extraction_utils import BatchFeature, FeatureExtractionMixin
+from ...file_utils import TensorType
+from ...image_utils import ImageFeatureExtractionMixin, ImageInput, is_torch_tensor
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+
+class GLPNFeatureExtractor(FeatureExtractionMixin, ImageFeatureExtractionMixin):
+    r"""
+    Constructs a GLPN feature extractor.
+
+    This feature extractor inherits from [`FeatureExtractionMixin`] which contains most of the main methods. Users
+    should refer to this superclass for more information regarding those methods.
+
+    Args:
+        do_resize (`bool`, *optional*, defaults to `True`):
+            Whether to resize the input based on certain `size_divisor`.
+        size_divisor (`int` or `Tuple(int)`, *optional*, defaults to 32):
+            Make sure the input is divisible by this value. Only has an effect if `do_resize` is set to `True`.
+        resample (`int`, *optional*, defaults to `PIL.Image.BILINEAR`):
+            An optional resampling filter. This can be one of `PIL.Image.NEAREST`, `PIL.Image.BOX`,
+            `PIL.Image.BILINEAR`, `PIL.Image.HAMMING`, `PIL.Image.BICUBIC` or `PIL.Image.LANCZOS`. Only has an effect
+            if `do_resize` is set to `True`.
+        do_rescale (`bool`, *optional*, defaults to `True`):
+             Whether or not to apply the scaling factor (to make pixel values floats between 0. and 1.).
+    """
+
+    model_input_names = ["pixel_values"]
+
+    def __init__(self, do_resize=True, size_divisor=32, resample=Image.BILINEAR, do_rescale=True, **kwargs):
+        super().__init__(**kwargs)
+        self.do_resize = do_resize
+        self.size_divisor = size_divisor
+        self.resample = resample
+        self.do_rescale = do_rescale
+
+    def _resize(self, image, size_divisor, resample):
+        if not isinstance(image, Image.Image):
+            image = self.to_pil_image(image)
+
+        width, height = image.size
+        new_h, new_w = height // size_divisor * size_divisor, width // size_divisor * size_divisor
+
+        image = self.resize(image, size=(new_w, new_h), resample=resample)
+
+        return image
+
+    def __call__(
+        self, images: ImageInput, return_tensors: Optional[Union[str, TensorType]] = None, **kwargs
+    ) -> BatchFeature:
+        """
+        Main method to prepare for the model one or several image(s).
+
+        <Tip warning={true}>
+
+        NumPy arrays and PyTorch tensors are converted to PIL images when resizing, so the most efficient is to pass
+        PIL images.
+
+        </Tip>
+
+        Args:
+            images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
+                The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch
+                tensor. In case of a NumPy array/PyTorch tensor, each image should be of shape (C, H, W), where C is a
+                number of channels, H and W are image height and width.
+
+            return_tensors (`str` or [`~file_utils.TensorType`], *optional*, defaults to `'np'`):
+                If set, will return tensors of a particular framework. Acceptable values are:
+
+                - `'tf'`: Return TensorFlow `tf.constant` objects.
+                - `'pt'`: Return PyTorch `torch.Tensor` objects.
+                - `'np'`: Return NumPy `np.ndarray` objects.
+                - `'jax'`: Return JAX `jnp.ndarray` objects.
+
+        Returns:
+            [`BatchFeature`]: A [`BatchFeature`] with the following fields:
+
+            - **pixel_values** -- Pixel values to be fed to a model, of shape (batch_size, num_channels, height,
+              width).
+        """
+        # Input type checking for clearer error
+        valid_images = False
+
+        # Check that images has a valid type
+        if isinstance(images, (Image.Image, np.ndarray)) or is_torch_tensor(images):
+            valid_images = True
+        elif isinstance(images, (list, tuple)):
+            if len(images) == 0 or isinstance(images[0], (Image.Image, np.ndarray)) or is_torch_tensor(images[0]):
+                valid_images = True
+
+        if not valid_images:
+            raise ValueError(
+                "Images must of type `PIL.Image.Image`, `np.ndarray` or `torch.Tensor` (single example), "
+                "`List[PIL.Image.Image]`, `List[np.ndarray]` or `List[torch.Tensor]` (batch of examples)."
+            )
+
+        is_batched = bool(
+            isinstance(images, (list, tuple))
+            and (isinstance(images[0], (Image.Image, np.ndarray)) or is_torch_tensor(images[0]))
+        )
+
+        if not is_batched:
+            images = [images]
+
+        # transformations (resizing + rescaling)
+        if self.do_resize and self.size_divisor is not None:
+            images = [
+                self._resize(image=image, size_divisor=self.size_divisor, resample=self.resample) for image in images
+            ]
+        if self.do_rescale:
+            images = [self.to_numpy_array(image=image) for image in images]
+
+        # return as BatchFeature
+        data = {"pixel_values": images}
+        encoded_inputs = BatchFeature(data=data, tensor_type=return_tensors)
+
+        return encoded_inputs

src/transformers/utils/dummy_pt_objects.py

@@ -1866,6 +1866,30 @@ def load_tf_weights_in_funnel(*args, **kwargs):
     requires_backends(load_tf_weights_in_funnel, ["torch"])
 
 
+GLPN_PRETRAINED_MODEL_ARCHIVE_LIST = None
+
+
+class GLPNForDepthEstimation(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class GLPNModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class GLPNPreTrainedModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
 GPT2_PRETRAINED_MODEL_ARCHIVE_LIST = None
 
 

src/transformers/utils/dummy_vision_objects.py

@@ -52,6 +52,13 @@ class DetrFeatureExtractor(metaclass=DummyObject):
         requires_backends(self, ["vision"])
 
 
+class GLPNFeatureExtractor(metaclass=DummyObject):
+    _backends = ["vision"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["vision"])
+
+
 class ImageGPTFeatureExtractor(metaclass=DummyObject):
     _backends = ["vision"]