Add MAE (huggingface#15120)

* First draft * More improvements * More improvements * More improvements * Fix embeddings * Add conversion script * Finish conversion script * More improvements * Fix forward pass * Remove print statements * Add weights initialization * Add initialization of decoder weights * Add support for other models in the conversion script * Fix patch_size for huge model * Fix most of the tests * Fix integration test * Fix docs * Fix archive_list * Apply suggestions from code review * Improve documentation * Apply more suggestions * Skip some tests due to non-deterministic behaviour * Fix test_initialization * Remove unneccessary initialization of nn.Embedding * Improve docs * Fix dummies * Remove ViTMAEFeatureExtractor from docs * Add model to README and table of contents * Delete inference file
avain · Jan 18, 2022 · 74bec98 · 74bec98
1 parent 2ae3be5
commit 74bec98
Show file tree

Hide file tree

Showing 17 changed files with 1,882 additions and 0 deletions.
diff --git a/README.md b/README.md
@@ -312,6 +312,7 @@ Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
 1. **[UniSpeechSat](https://huggingface.co/docs/transformers/model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER
 AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
+1. **[ViTMAE)](https://huggingface.co/docs/transformers/master/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[WavLM](https://huggingface.co/docs/transformers/master/model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.

diff --git a/README_ko.md b/README_ko.md
@@ -291,6 +291,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[UniSpeechSat](https://huggingface.co/docs/transformers/model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
+1. **[ViTMAE)](https://huggingface.co/docs/transformers/master/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2Phoneme](https://huggingface.co/docs/master/transformers/model_doc/wav2vec2_phoneme)** (from Facebook AI) released with the paper [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
 1. **[WavLM](https://huggingface.co/docs/transformers/master/model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.

diff --git a/README_zh-hans.md b/README_zh-hans.md
@@ -315,6 +315,7 @@ conda install -c huggingface transformers
 1. **[UniSpeechSat](https://huggingface.co/docs/transformers/model_doc/unispeech-sat)** (来自 Microsoft Research) 伴随论文 [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) 由 Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu 发布。
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (来自 Google AI) 伴随论文 [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) 由 Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby 发布。
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (来自 UCLA NLP) 伴随论文 [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) 由 Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang 发布。
+1. **[ViTMAE)](https://huggingface.co/docs/transformers/master/model_doc/vit_mae)** (来自 Meta AI) 伴随论文 [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) 由 Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick 发布。
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (来自 Facebook AI) 伴随论文 [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) 由 Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli 发布。
 1. **[Wav2Vec2Phoneme](https://huggingface.co/docs/master/transformers/model_doc/wav2vec2_phoneme)** (来自 Facebook AI) 伴随论文 [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) 由 Qiantong Xu, Alexei Baevski, Michael Auli 发布。
 1. **[WavLM](https://huggingface.co/docs/transformers/master/model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.

diff --git a/README_zh-hant.md b/README_zh-hant.md
@@ -327,6 +327,7 @@ conda install -c huggingface transformers
 1. **[UniSpeechSat](https://huggingface.co/docs/transformers/model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby. 
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
+1. **[ViTMAE)](https://huggingface.co/docs/transformers/master/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2Phoneme](https://huggingface.co/docs/master/transformers/model_doc/wav2vec2_phoneme)** (from Facebook AI) released with the paper [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
 1. **[WavLM](https://huggingface.co/docs/transformers/master/model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.

diff --git a/docs/source/_toctree.yml b/docs/source/_toctree.yml
@@ -288,6 +288,8 @@
       title: Vision Text Dual Encoder
     - local: model_doc/vit
       title: Vision Transformer (ViT)
+    - local: model_doc/vit_mae
+      title: ViTMAE
     - local: model_doc/visual_bert
       title: VisualBERT
     - local: model_doc/wav2vec2

diff --git a/docs/source/index.mdx b/docs/source/index.mdx
@@ -171,6 +171,7 @@ conversion utilities for the following models.
 1. **[UniSpeech](model_doc/unispeech)** (from Microsoft Research) released with the paper [UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data](https://arxiv.org/abs/2101.07597) by Chengyi Wang, Yu Wu, Yao Qian, Kenichi Kumatani, Shujie Liu, Furu Wei, Michael Zeng, Xuedong Huang.
 1. **[UniSpeechSat](model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
 1. **[Vision Transformer (ViT)](model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
+1. **[ViTMAE)](model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
 1. **[VisualBERT](model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[WavLM](model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.
 1. **[Wav2Vec2](model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
@@ -269,6 +270,7 @@ Flax), PyTorch, and/or TensorFlow.
 |    VisionTextDualEncoder    |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
 |         VisualBert          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             ViT             |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
+|           ViTMAE            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          Wav2Vec2           |       ✅       |       ❌       |       ✅        |         ✅         |      ✅      |
 |            WavLM            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             XLM             |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |

diff --git a/docs/source/model_doc/vit_mae.mdx b/docs/source/model_doc/vit_mae.mdx
@@ -0,0 +1,63 @@
+<!--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.
+-->
+
+# ViTMAE
+
+## Overview
+
+The ViTMAE model was proposed in [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377v2) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li,
+Piotr Dollár, Ross Girshick. The paper shows that, by pre-training a Vision Transformer (ViT) to reconstruct pixel values for masked patches, one can get results after
+fine-tuning that outperform supervised pre-training.
+
+The abstract from the paper is the following:
+
+*This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the
+input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates
+only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask
+tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs
+enables us to train large models efficiently and effectively: we accelerate training (by 3x or more) and improve accuracy. Our scalable approach allows for learning high-capacity
+models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream
+tasks outperforms supervised pre-training and shows promising scaling behavior.*
+
+Tips:
+
+- MAE (masked auto encoding) is a method for self-supervised pre-training of Vision Transformers (ViTs). The pre-training objective is relatively simple:
+by masking a large portion (75%) of the image patches, the model must reconstruct raw pixel values. One can use [`ViTMAEForPreTraining`] for this purpose.
+- After pre-training, one "throws away" the decoder used to reconstruct pixels, and one uses the encoder for fine-tuning/linear probing. This means that after
+fine-tuning, one can directly plug in the weights into a [`ViTForImageClassification`].
+- One can use [`ViTFeatureExtractor`] to prepare images for the model. See the code examples for more info. 
+- Note that the encoder of MAE is only used to encode the visual patches. The encoded patches are then concatenated with mask tokens, which the decoder (which also
+consists of Transformer blocks) takes as input. Each mask token is a shared, learned vector that indicates the presence of a missing patch to be predicted. Fixed
+sin/cos position embeddings are added both to the input of the encoder and the decoder.
+
+<img src="https://user-images.githubusercontent.com/11435359/146857310-f258c86c-fde6-48e8-9cee-badd2b21bd2c.png"
+alt="drawing" width="600"/> 
+
+<small> MAE architecture. Taken from the <a href="https://arxiv.org/abs/2111.06377">original paper.</a> </small>
+
+This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code can be found [here](https://github.com/facebookresearch/mae).
+
+## ViTMAEConfig
+
+[[autodoc]] ViTMAEConfig
+
+
+## ViTMAEModel
+
+[[autodoc]] ViTMAEModel
+    - forward
+
+
+## ViTMAEForPreTraining
+
+[[autodoc]] transformers.ViTMAEForPreTraining
+    - forward
diff --git a/src/transformers/__init__.py b/src/transformers/__init__.py
@@ -312,6 +312,7 @@
     "models.vision_text_dual_encoder": ["VisionTextDualEncoderConfig", "VisionTextDualEncoderProcessor"],
     "models.visual_bert": ["VISUAL_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "VisualBertConfig"],
     "models.vit": ["VIT_PRETRAINED_CONFIG_ARCHIVE_MAP", "ViTConfig"],
+    "models.vit_mae": ["VIT_MAE_PRETRAINED_CONFIG_ARCHIVE_MAP", "ViTMAEConfig"],
     "models.wav2vec2": [
         "WAV_2_VEC_2_PRETRAINED_CONFIG_ARCHIVE_MAP",
         "Wav2Vec2Config",
@@ -627,6 +628,7 @@
     _import_structure["modeling_utils"] = ["Conv1D", "PreTrainedModel", "apply_chunking_to_forward", "prune_layer"]
 
     # PyTorch models structure
+
     _import_structure["models.albert"].extend(
         [
             "ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST",
@@ -1402,6 +1404,15 @@
             "ViTPreTrainedModel",
         ]
     )
+    _import_structure["models.vit_mae"].extend(
+        [
+            "VIT_MAE_PRETRAINED_MODEL_ARCHIVE_LIST",
+            "ViTMAEForPreTraining",
+            "ViTMAELayer",
+            "ViTMAEModel",
+            "ViTMAEPreTrainedModel",
+        ]
+    )
     _import_structure["models.wav2vec2"].extend(
         [
             "WAV_2_VEC_2_PRETRAINED_MODEL_ARCHIVE_LIST",
@@ -2401,6 +2412,7 @@
     from .models.vision_text_dual_encoder import VisionTextDualEncoderConfig, VisionTextDualEncoderProcessor
     from .models.visual_bert import VISUAL_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, VisualBertConfig
     from .models.vit import VIT_PRETRAINED_CONFIG_ARCHIVE_MAP, ViTConfig
+    from .models.vit_mae import VIT_MAE_PRETRAINED_CONFIG_ARCHIVE_MAP, ViTMAEConfig
     from .models.wav2vec2 import (
         WAV_2_VEC_2_PRETRAINED_CONFIG_ARCHIVE_MAP,
         Wav2Vec2Config,
@@ -3307,6 +3319,13 @@
             ViTModel,
             ViTPreTrainedModel,
         )
+        from .models.vit_mae import (
+            VIT_MAE_PRETRAINED_MODEL_ARCHIVE_LIST,
+            ViTMAEForPreTraining,
+            ViTMAELayer,
+            ViTMAEModel,
+            ViTMAEPreTrainedModel,
+        )
         from .models.wav2vec2 import (
             WAV_2_VEC_2_PRETRAINED_MODEL_ARCHIVE_LIST,
             Wav2Vec2ForAudioFrameClassification,

diff --git a/src/transformers/models/__init__.py b/src/transformers/models/__init__.py
@@ -108,6 +108,7 @@
     vision_text_dual_encoder,
     visual_bert,
     vit,
+    vit_mae,
     wav2vec2,
     wav2vec2_phoneme,
     wav2vec2_with_lm,