Add AltCLIP (#20446)

* add altclip * update * fix wrong title * fix the copyright in readme * add altclip model * add altclip * fix test_gradient_checkpointing_enable_disable * code * add return class * add projection_state * "fix pretrained model bug" * delete print and fix 2 test instances. * delete token * rm xlmr * one model one file. * empty commit to trigger CI * Fix modeling_outputs.py * Fix __init__ * Fix quality * Fix modeling file docstring * Fix README.md * Fix test file * add vision model * empty commit to trigger CI * fix * fix * fix * fix * fix * fix * fix * fix * fix * del token in mdx file * fix * fix * fix * remove altrob from test list * add vision test * fix fx * fix * fix * fix * trigger CI * fix copies * fix tests * fix style * fix quality * update * recover import * recover * add , * recover * fix copies * trigger CI * fix * some of review * update * remove import * last 2 * fix * fix style * fix style * fix bug * fix uncomment * fix * update * fix * second review * empty commit to trigger CI * empty commit to trigger CI * fix position * fix * empty commit to trigger CI * empty commit to trigger CI * third comment * Update docs/source/en/model_doc/altclip.mdx Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com> * Update docs/source/en/model_doc/altclip.mdx Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com> * Update src/transformers/__init__.py Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com> * Update src/transformers/models/altclip/configuration_altclip.py Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com> * Update src/transformers/models/altclip/modeling_altclip.py Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com> * Update src/transformers/models/altclip/processing_altclip.py Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com> * Update src/transformers/models/altclip/modeling_altclip.py Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com> * fix merge * fix copies * update * update * empty commit to trigger CI * fix code example * empty commit to trigger CI * fix * empty commit to trigger CI * empty commit to trigger CI Co-authored-by: shunxing1234 <xw747777271@gmail.com> Co-authored-by: ydshieh <ydshieh@users.noreply.github.com> Co-authored-by: shunxing1234 <33774367+shunxing1234@users.noreply.github.com> Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>
2023-01-04 16:18:57 +08:00
parent 45da7cec5a
commit ce85686a1f
25 changed files with 3044 additions and 1 deletions
--- a/README.md
+++ b/README.md
@@ -263,6 +263,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 🤗 Transformers currently provides the following architectures (see [here](https://huggingface.co/docs/transformers/model_summary) for a high-level summary of each them):

 1. **[ALBERT](https://huggingface.co/docs/transformers/model_doc/albert)** (from Google Research and the Toyota Technological Institute at Chicago) released with the paper [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://arxiv.org/abs/1909.11942), by Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut.
+1. **[AltCLIP](https://huggingface.co/docs/transformers/main/model_doc/altclip)** (from BAAI) released with the paper [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679) by Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell.
 1. **[Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer)** (from MIT) released with the paper [AST: Audio Spectrogram Transformer](https://arxiv.org/abs/2104.01778) by Yuan Gong, Yu-An Chung, James Glass.
 1. **[BART](https://huggingface.co/docs/transformers/model_doc/bart)** (from Facebook) released with the paper [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/abs/1910.13461) by Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer.
 1. **[BARThez](https://huggingface.co/docs/transformers/model_doc/barthez)** (from École polytechnique) released with the paper [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis.
--- a/README_es.md
+++ b/README_es.md
@@ -263,6 +263,7 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt
 🤗 Transformers actualmente proporciona las siguientes arquitecturas (ver [aquí](https://huggingface.co/docs/transformers/model_summary) para un resumen de alto nivel de cada uno de ellas.):

 1. **[ALBERT](https://huggingface.co/docs/transformers/model_doc/albert)** (from Google Research and the Toyota Technological Institute at Chicago) released with the paper [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://arxiv.org/abs/1909.11942), by Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut.
+1. **[AltCLIP](https://huggingface.co/docs/transformers/main/model_doc/altclip)** (from BAAI) released with the paper [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679) by Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell.
 1. **[Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer)** (from MIT) released with the paper [AST: Audio Spectrogram Transformer](https://arxiv.org/abs/2104.01778) by Yuan Gong, Yu-An Chung, James Glass.
 1. **[BART](https://huggingface.co/docs/transformers/model_doc/bart)** (from Facebook) released with the paper [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/abs/1910.13461) by Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer.
 1. **[BARThez](https://huggingface.co/docs/transformers/model_doc/barthez)** (from École polytechnique) released with the paper [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis.
--- a/README_hd.md
+++ b/README_hd.md
@@ -236,6 +236,7 @@ conda install -c huggingface transformers
 🤗 ट्रांसफॉर्मर वर्तमान में निम्नलिखित आर्किटेक्चर का समर्थन करते हैं (मॉडल के अवलोकन के लिए [यहां] देखें (https://huggingface.co/docs/transformers/model_summary))：

 1. **[ALBERT](https://huggingface.co/docs/transformers/model_doc/albert)** (Google Research and the Toyota Technological Institute at Chicago) साथ थीसिस [ALBERT: A Lite BERT for Self-supervised भाषा प्रतिनिधित्व सीखना](https://arxiv.org/abs/1909.11942), झेंझोंग लैन, मिंगदा चेन, सेबेस्टियन गुडमैन, केविन गिम्पेल, पीयूष शर्मा, राडू सोरिकट
+1. **[AltCLIP](https://huggingface.co/docs/transformers/main/model_doc/altclip)** (from BAAI) released with the paper [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679) by Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell.
 1. **[Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer)** (from MIT) released with the paper [AST: Audio Spectrogram Transformer](https://arxiv.org/abs/2104.01778) by Yuan Gong, Yu-An Chung, James Glass.
 1. **[BART](https://huggingface.co/docs/transformers/model_doc/bart)** (फेसबुक) साथ थीसिस [बार्ट: प्राकृतिक भाषा निर्माण, अनुवाद के लिए अनुक्रम-से-अनुक्रम पूर्व प्रशिक्षण , और समझ] (https://arxiv.org/pdf/1910.13461.pdf) पर निर्भर माइक लुईस, यिनहान लियू, नमन गोयल, मार्जन ग़ज़विनिनेजाद, अब्देलरहमान मोहम्मद, ओमर लेवी, वेस स्टोयानोव और ल्यूक ज़ेटलमॉयर
 1. **[BARThez](https://huggingface.co/docs/transformers/model_doc/barthez)** (से École polytechnique) साथ थीसिस [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) पर निर्भर Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis रिहाई।
--- a/README_ja.md
+++ b/README_ja.md
@@ -298,6 +298,7 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ
 🤗Transformersは現在、以下のアーキテクチャを提供しています（それぞれのハイレベルな要約は[こちら](https://huggingface.co/docs/transformers/model_summary)を参照してください）:

 1. **[ALBERT](https://huggingface.co/docs/transformers/model_doc/albert)** (Google Research and the Toyota Technological Institute at Chicago から) Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut から公開された研究論文: [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://arxiv.org/abs/1909.11942)
+1. **[AltCLIP](https://huggingface.co/docs/transformers/main/model_doc/altclip)** (BAAI から) Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell から公開された研究論文: [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679)
 1. **[Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer)** (MIT から) Yuan Gong, Yu-An Chung, James Glass から公開された研究論文: [AST: Audio Spectrogram Transformer](https://arxiv.org/abs/2104.01778)
 1. **[BART](https://huggingface.co/docs/transformers/model_doc/bart)** (Facebook から) Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer から公開された研究論文: [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/abs/1910.13461)
 1. **[BARThez](https://huggingface.co/docs/transformers/model_doc/barthez)** (École polytechnique から) Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis から公開された研究論文: [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321)
--- a/README_ko.md
+++ b/README_ko.md
@@ -213,6 +213,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 🤗 Transformers는 다음 모델들을 제공합니다 (각 모델의 요약은 [여기](https://huggingface.co/docs/transformers/model_summary)서 확인하세요):

 1. **[ALBERT](https://huggingface.co/docs/transformers/model_doc/albert)** (from Google Research and the Toyota Technological Institute at Chicago) released with the paper [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://arxiv.org/abs/1909.11942), by Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut.
+1. **[AltCLIP](https://huggingface.co/docs/transformers/main/model_doc/altclip)** (from BAAI) released with the paper [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679) by Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell.
 1. **[Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer)** (from MIT) released with the paper [AST: Audio Spectrogram Transformer](https://arxiv.org/abs/2104.01778) by Yuan Gong, Yu-An Chung, James Glass.
 1. **[BART](https://huggingface.co/docs/transformers/model_doc/bart)** (from Facebook) released with the paper [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/pdf/1910.13461.pdf) by Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer.
 1. **[BARThez](https://huggingface.co/docs/transformers/model_doc/barthez)** (from École polytechnique) released with the paper [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis.
--- a/README_zh-hans.md
+++ b/README_zh-hans.md
@@ -237,6 +237,7 @@ conda install -c huggingface transformers
 🤗 Transformers 目前支持如下的架构（模型概述请阅[这里](https://huggingface.co/docs/transformers/model_summary)）：

 1. **[ALBERT](https://huggingface.co/docs/transformers/model_doc/albert)** (来自 Google Research and the Toyota Technological Institute at Chicago) 伴随论文 [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://arxiv.org/abs/1909.11942), 由 Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut 发布。
+1. **[AltCLIP](https://huggingface.co/docs/transformers/main/model_doc/altclip)** (来自 BAAI) 伴随论文 [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679) 由 Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell 发布。
 1. **[Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer)** (来自 MIT) 伴随论文 [AST: Audio Spectrogram Transformer](https://arxiv.org/abs/2104.01778) 由 Yuan Gong, Yu-An Chung, James Glass 发布。
 1. **[BART](https://huggingface.co/docs/transformers/model_doc/bart)** (来自 Facebook) 伴随论文 [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/pdf/1910.13461.pdf) 由 Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer 发布。
 1. **[BARThez](https://huggingface.co/docs/transformers/model_doc/barthez)** (来自 École polytechnique) 伴随论文 [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) 由 Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis 发布。
--- a/README_zh-hant.md
+++ b/README_zh-hant.md
@@ -249,6 +249,7 @@ conda install -c huggingface transformers
 🤗 Transformers 目前支援以下的架構（模型概覽請參閱[這裡](https://huggingface.co/docs/transformers/model_summary)）：

 1. **[ALBERT](https://huggingface.co/docs/transformers/model_doc/albert)** (from Google Research and the Toyota Technological Institute at Chicago) released with the paper [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://arxiv.org/abs/1909.11942), by Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut.
+1. **[AltCLIP](https://huggingface.co/docs/transformers/main/model_doc/altclip)** (from BAAI) released with the paper [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679) by Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell.
 1. **[Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer)** (from MIT) released with the paper [AST: Audio Spectrogram Transformer](https://arxiv.org/abs/2104.01778) by Yuan Gong, Yu-An Chung, James Glass.
 1. **[BART](https://huggingface.co/docs/transformers/model_doc/bart)** (from Facebook) released with the paper [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/pdf/1910.13461.pdf) by Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer.
 1. **[BARThez](https://huggingface.co/docs/transformers/model_doc/barthez)** (from École polytechnique) released with the paper [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis.
--- a/docs/source/en/_toctree.yml
+++ b/docs/source/en/_toctree.yml
@@ -498,6 +498,8 @@
      title: Audio models
    - isExpanded: false
      sections:
+      - local: model_doc/altclip
+        title: AltCLIP
      - local: model_doc/blip
        title: BLIP
      - local: model_doc/chinese_clip
--- a/docs/source/en/index.mdx
+++ b/docs/source/en/index.mdx
@@ -50,6 +50,7 @@ The documentation is organized into five sections:
 <!--This list is updated automatically from the README with _make fix-copies_. Do not update manually! -->

 1. **[ALBERT](model_doc/albert)** (from Google Research and the Toyota Technological Institute at Chicago) released with the paper [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://arxiv.org/abs/1909.11942), by Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut.
+1. **[AltCLIP](model_doc/altclip)** (from BAAI) released with the paper [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679) by Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell.
 1. **[Audio Spectrogram Transformer](model_doc/audio-spectrogram-transformer)** (from MIT) released with the paper [AST: Audio Spectrogram Transformer](https://arxiv.org/abs/2104.01778) by Yuan Gong, Yu-An Chung, James Glass.
 1. **[BART](model_doc/bart)** (from Facebook) released with the paper [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/abs/1910.13461) by Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer.
 1. **[BARThez](model_doc/barthez)** (from École polytechnique) released with the paper [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis.
@@ -230,6 +231,7 @@ Flax), PyTorch, and/or TensorFlow.
 |             Model             | Tokenizer slow | Tokenizer fast | PyTorch support | TensorFlow support | Flax Support |
 |:-----------------------------:|:--------------:|:--------------:|:---------------:|:------------------:|:------------:|
 |            ALBERT             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|            AltCLIP            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 | Audio Spectrogram Transformer |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             BART              |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |             BEiT              |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
--- a/docs/source/en/model_doc/altclip.mdx
+++ b/docs/source/en/model_doc/altclip.mdx
@@ -0,0 +1,107 @@
+<!--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.
+-->
+
+# AltCLIP
+
+## Overview
+
+The AltCLIP model was proposed in [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679v2) by Zhongzhi Chen, Guang Liu, Bo-Wen Zhang, Fulong Ye, Qinghong Yang, Ledell Wu. AltCLIP
+(Altering the Language Encoder in CLIP) is a neural network trained on a variety of image-text and text-text pairs. By switching CLIP's
+text encoder with a pretrained multilingual text encoder XLM-R, we could obtain very close performances with CLIP on almost all tasks, and extended original CLIP's capabilities such as multilingual understanding.
+
+The abstract from the paper is the following:
+
+*In this work, we present a conceptually simple and effective method to train a strong bilingual multimodal representation model. 
+Starting from the pretrained multimodal representation model CLIP released by OpenAI, we switched its text encoder with a pretrained 
+multilingual text encoder XLM-R, and aligned both languages and image representations by a two-stage training schema consisting of 
+teacher learning and contrastive learning. We validate our method through evaluations of a wide range of tasks. We set new state-of-the-art 
+performances on a bunch of tasks including ImageNet-CN, Flicker30k- CN, and COCO-CN. Further, we obtain very close performances with 
+CLIP on almost all tasks, suggesting that one can simply alter the text encoder in CLIP for extended capabilities such as multilingual understanding.*
+
+## Usage
+
+The usage of AltCLIP is very similar to the CLIP. the difference between CLIP is the text encoder. Note that we use bidirectional attention instead of casual attention
+and we take the [CLS] token in XLM-R to represent text embedding.
+
+AltCLIP is a multi-modal vision and language model. It can be used for image-text similarity and for zero-shot image
+classification. AltCLIP uses a ViT like transformer to get visual features and a bidirectional language model to get the text
+features. Both the text and visual features are then projected to a latent space with identical dimension. The dot
+product between the projected image and text features is then used as a similar score.
+
+To feed images to the Transformer encoder, each image is split into a sequence of fixed-size non-overlapping patches,
+which are then linearly embedded. A [CLS] token is added to serve as representation of an entire image. The authors
+also add absolute position embeddings, and feed the resulting sequence of vectors to a standard Transformer encoder.
+The [`CLIPImageProcessor`] can be used to resize (or rescale) and normalize images for the model.
+
+The [`AltCLIPProcessor`] wraps a [`CLIPImageProcessor`] and a [`XLMRobertaTokenizer`] into a single instance to both
+encode the text and prepare the images. The following example shows how to get the image-text similarity scores using
+[`AltCLIPProcessor`] and [`AltCLIPModel`].
+
+
+```python
+>>> from PIL import Image
+>>> import requests
+
+>>> from transformers import AltCLIPModel, AltCLIPProcessor
+
+>>> model = AltCLIPModel.from_pretrained("BAAI/AltCLIP")
+>>> processor = AltCLIPProcessor.from_pretrained("BAAI/AltCLIP")
+
+>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
+>>> image = Image.open(requests.get(url, stream=True).raw)
+
+>>> inputs = processor(text=["a photo of a cat", "a photo of a dog"], images=image, return_tensors="pt", padding=True)
+
+>>> outputs = model(**inputs)
+>>> logits_per_image = outputs.logits_per_image  # this is the image-text similarity score
+>>> probs = logits_per_image.softmax(dim=1)  # we can take the softmax to get the label probabilities
+```
+
+Tips:
+
+This model is build on `CLIPModel`, so use it like a original CLIP. 
+
+This model was contributed by [jongjyh](https://huggingface.co/jongjyh).
+
+## AltCLIPConfig
+
+[[autodoc]] AltCLIPConfig
+    - from_text_vision_configs
+
+## AltCLIPTextConfig
+
+[[autodoc]] AltCLIPTextConfig
+
+## AltCLIPVisionConfig
+
+[[autodoc]] AltCLIPVisionConfig
+
+## AltCLIPProcessor
+
+[[autodoc]] AltCLIPProcessor
+
+## AltCLIPModel
+
+[[autodoc]] AltCLIPModel
+    - forward
+    - get_text_features
+    - get_image_features
+
+## AltCLIPTextModel
+
+[[autodoc]] AltCLIPTextModel
+    - forward
+
+## AltCLIPVisionModel
+
+[[autodoc]] AltCLIPVisionModel
+    - forward
--- a/src/transformers/init.py
+++ b/src/transformers/init.py
@@ -124,6 +124,13 @@ _import_structure = {
    "models": [],
    # Models
    "models.albert": ["ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "AlbertConfig"],
+    "models.altclip": [
+        "ALTCLIP_PRETRAINED_CONFIG_ARCHIVE_MAP",
+        "AltCLIPConfig",
+        "AltCLIPProcessor",
+        "AltCLIPTextConfig",
+        "AltCLIPVisionConfig",
+    ],
    "models.audio_spectrogram_transformer": [
        "AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP",
        "ASTConfig",
@@ -921,7 +928,15 @@ else:
            "load_tf_weights_in_albert",
        ]
    )
-
+    _import_structure["models.altclip"].extend(
+        [
+            "ALTCLIP_PRETRAINED_MODEL_ARCHIVE_LIST",
+            "AltCLIPModel",
+            "AltCLIPPreTrainedModel",
+            "AltCLIPTextModel",
+            "AltCLIPVisionModel",
+        ]
+    )
    _import_structure["models.audio_spectrogram_transformer"].extend(
        [
            "AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST",
@@ -3482,6 +3497,13 @@ if TYPE_CHECKING:
        load_tf2_weights_in_pytorch_model,
    )
    from .models.albert import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, AlbertConfig
+    from .models.altclip import (
+        ALTCLIP_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        AltCLIPConfig,
+        AltCLIPProcessor,
+        AltCLIPTextConfig,
+        AltCLIPVisionConfig,
+    )
    from .models.audio_spectrogram_transformer import (
        AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP,
        ASTConfig,
@@ -4181,6 +4203,13 @@ if TYPE_CHECKING:
            AlbertPreTrainedModel,
            load_tf_weights_in_albert,
        )
+        from .models.altclip import (
+            ALTCLIP_PRETRAINED_MODEL_ARCHIVE_LIST,
+            AltCLIPModel,
+            AltCLIPPreTrainedModel,
+            AltCLIPTextModel,
+            AltCLIPVisionModel,
+        )
        from .models.audio_spectrogram_transformer import (
            AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
            ASTForAudioClassification,
--- a/src/transformers/modeling_outputs.py
+++ b/src/transformers/modeling_outputs.py
@@ -1318,3 +1318,40 @@ class BackboneOutput(ModelOutput):
    feature_maps: Tuple[torch.FloatTensor] = None
    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
    attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+@dataclass
+class BaseModelOutputWithPoolingAndProjection(ModelOutput):
+    """
+    Base class for model's outputs that also contains a pooling of the last hidden states.
+
+    Args:
+        last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
+            Sequence of hidden-states at the output of the last layer of the model.
+        pooler_output (`torch.FloatTensor` of shape `(batch_size, hidden_size)`):
+            Last layer hidden-state of the first token of the sequence (classification token) after further processing
+            through the layers used for the auxiliary pretraining task. E.g. for BERT-family of models, this returns
+            the classification token after processing through a linear layer and a tanh activation function. The linear
+            layer weights are trained from the next sentence prediction (classification) objective during pretraining.
+        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, if the model has an embedding layer, +
+            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
+
+            Hidden-states of the model at the output of each layer plus the optional 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, sequence_length,
+            sequence_length)`.
+
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
+        projection_state (`tuple(torch.FloatTensor)`, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` of shape `(batch_size,config.project_dim)`.
+
+            Text embeddings before the projection layer, used to mimic the last hidden state of the teacher encoder.
+    """
+
+    last_hidden_state: torch.FloatTensor = None
+    pooler_output: torch.FloatTensor = None
+    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    attentions: Optional[Tuple[torch.FloatTensor]] = None
+    projection_state: Optional[Tuple[torch.FloatTensor]] = None
--- a/src/transformers/models/init.py
+++ b/src/transformers/models/init.py
@@ -18,6 +18,7 @@

 from . import (
    albert,
+    altclip,
    audio_spectrogram_transformer,
    auto,
    bart,
--- a/src/transformers/models/altclip/init.py
+++ b/src/transformers/models/altclip/init.py
@@ -0,0 +1,76 @@
+# 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 2020 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 ...utils import OptionalDependencyNotAvailable, _LazyModule, is_tokenizers_available, is_torch_available
+
+
+_import_structure = {
+    "configuration_altclip": [
+        "ALTCLIP_PRETRAINED_CONFIG_ARCHIVE_MAP",
+        "AltCLIPConfig",
+        "AltCLIPTextConfig",
+        "AltCLIPVisionConfig",
+    ],
+    "processing_altclip": ["AltCLIPProcessor"],
+}
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_altclip"] = [
+        "ALTCLIP_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "AltCLIPPreTrainedModel",
+        "AltCLIPModel",
+        "AltCLIPTextModel",
+        "AltCLIPVisionModel",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_altclip import (
+        ALTCLIP_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        AltCLIPConfig,
+        AltCLIPTextConfig,
+        AltCLIPVisionConfig,
+    )
+    from .processing_altclip import AltCLIPProcessor
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_altclip import (
+            ALTCLIP_PRETRAINED_MODEL_ARCHIVE_LIST,
+            AltCLIPModel,
+            AltCLIPPreTrainedModel,
+            AltCLIPTextModel,
+            AltCLIPVisionModel,
+        )
+
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
--- a/src/transformers/models/altclip/configuration_altclip.py
+++ b/src/transformers/models/altclip/configuration_altclip.py
@@ -0,0 +1,357 @@
+# coding=utf-8
+# Copyright 2022 WenXiang ZhongzhiCheng LedellWu LiuGuang BoWenZhang 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.
+""" AltCLIP model configuration"""
+import copy
+import os
+from typing import Union
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+ALTCLIP_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "BAAI/AltCLIP": "https://huggingface.co/BAAI/AltCLIP/resolve/main/config.json",
+    # See all AltCLIP models at https://huggingface.co/models?filter=altclip
+}
+
+
+class AltCLIPTextConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`AltCLIPTextModel`]. It is used to instantiate a
+    AltCLIP text 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 AltCLIP
+    [BAAI/AltCLIP](https://huggingface.co/BAAI/AltCLIP) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 250002):
+            Vocabulary size of the AltCLIP model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`AltCLIPTextModel`].
+        hidden_size (`int`, *optional*, defaults to 1024):
+            Dimensionality of the encoder layers and the pooler layer.
+        num_hidden_layers (`int`, *optional*, defaults to 24):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 16):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        intermediate_size (`int`, *optional*, defaults to 4096):
+            Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
+        hidden_act (`str` or `Callable`, *optional*, defaults to `"gelu"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"silu"` and `"gelu_new"` are supported.
+        hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
+            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
+        attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
+            The dropout ratio for the attention probabilities.
+        max_position_embeddings (`int`, *optional*, defaults to 514):
+            The maximum sequence length that this model might ever be used with. Typically set this to something large
+            just in case (e.g., 512 or 1024 or 2048).
+        type_vocab_size (`int`, *optional*, defaults to 2):
+            The vocabulary size of the `token_type_ids` passed when calling [`AltCLIPTextModel`]
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-5):
+            The epsilon used by the layer normalization layers.
+        position_embedding_type (`str`, *optional*, defaults to `"absolute"`):
+            Type of position embedding. Choose one of `"absolute"`, `"relative_key"`, `"relative_key_query"`. For
+            positional embeddings use `"absolute"`. For more information on `"relative_key"`, please refer to
+            [Self-Attention with Relative Position Representations (Shaw et al.)](https://arxiv.org/abs/1803.02155).
+            For more information on `"relative_key_query"`, please refer to *Method 4* in [Improve Transformer Models
+            with Better Relative Position Embeddings (Huang et al.)](https://arxiv.org/abs/2009.13658).
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        classifier_dropout (`float`, *optional*):
+            The dropout ratio for the classification head.
+        project_dim (`int`, *optional*, defaults to 768):
+            The dimentions of the teacher model before the mapping layer.
+        pooler_fn (`str`, *optional*, defaults to `"cls"`):
+            Type of pooler we use. We take the first token as pooled output.
+
+    Examples:
+
+    ```python
+    >>> from transformers import AltCLIPTextModel, AltCLIPTextConfig
+
+    >>> # Initializing a AltCLIPTextConfig with BAAI/AltCLIP style configuration
+    >>> configuration = AltCLIPTextConfig()
+
+    >>> # Initializing a AltCLIPTextModel (with random weights) from the BAAI/AltCLIP style configuration
+    >>> model = AltCLIPTextModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+    model_type = "altclip_text_model"
+
+    def __init__(
+        self,
+        vocab_size=250002,
+        hidden_size=1024,
+        num_hidden_layers=24,
+        num_attention_heads=16,
+        intermediate_size=4096,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=514,
+        type_vocab_size=1,
+        initializer_range=0.02,
+        initializer_factor=0.02,
+        layer_norm_eps=1e-05,
+        pad_token_id=1,
+        bos_token_id=0,
+        eos_token_id=2,
+        position_embedding_type="absolute",
+        use_cache=True,
+        classifier_dropout=None,
+        project_dim=768,
+        pooler_fn="cls",
+        **kwargs
+    ):
+        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
+
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.hidden_act = hidden_act
+        self.intermediate_size = intermediate_size
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.initializer_range = initializer_range
+        self.initializer_factor = initializer_factor
+        self.layer_norm_eps = layer_norm_eps
+        self.position_embedding_type = position_embedding_type
+        self.use_cache = use_cache
+        self.classifier_dropout = classifier_dropout
+        self.project_dim = project_dim
+        self.pooler_fn = pooler_fn
+
+
+class AltCLIPVisionConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`AltCLIPModel`]. It is used to instantiate an
+    AltCLIP 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 AltCLIP
+    [BAAI/AltCLIP](https://huggingface.co/BAAI/AltCLIP) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        hidden_size (`int`, *optional*, defaults to 768):
+            Dimensionality of the encoder layers and the pooler layer.
+        intermediate_size (`int`, *optional*, defaults to 3072):
+            Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
+        num_hidden_layers (`int`, *optional*, defaults to 12):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 12):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        image_size (`int`, *optional*, defaults to 224):
+            The size (resolution) of each image.
+        patch_size (`int`, *optional*, defaults to 32):
+            The size (resolution) of each patch.
+        hidden_act (`str` or `function`, *optional*, defaults to `"quick_gelu"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported. layer_norm_eps (`float`, *optional*,
+            defaults to 1e-5): The epsilon used by the layer normalization layers.
+        dropout (`float`, *optional*, defaults to 0.0):
+            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
+        attention_dropout (`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.
+        initializer_factor (`float``, *optional*, defaults to 1):
+            A factor for initializing all weight matrices (should be kept to 1, used internally for initialization
+            testing).
+
+    Example:
+
+    ```python
+    >>> from transformers import AltCLIPVisionConfig, AltCLIPVisionModel
+
+    >>> # Initializing a AltCLIPVisionConfig with BAAI/AltCLIP style configuration
+    >>> configuration = AltCLIPVisionConfig()
+
+    >>> # Initializing a AltCLIPVisionModel (with random weights) from the BAAI/AltCLIP style configuration
+    >>> model = AltCLIPVisionModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "altclip_vision_model"
+
+    def __init__(
+        self,
+        hidden_size=768,
+        intermediate_size=3072,
+        projection_dim=512,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        num_channels=3,
+        image_size=224,
+        patch_size=32,
+        hidden_act="quick_gelu",
+        layer_norm_eps=0.00001,
+        dropout=0.0,
+        attention_dropout=0.0,
+        initializer_range=0.02,
+        initializer_factor=1.0,
+        **kwargs
+    ):
+        super().__init__(**kwargs)
+
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.projection_dim = projection_dim
+        self.dropout = dropout
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_channels = num_channels
+        self.patch_size = patch_size
+        self.image_size = image_size
+        self.initializer_range = initializer_range
+        self.initializer_factor = initializer_factor
+        self.attention_dropout = attention_dropout
+        self.layer_norm_eps = layer_norm_eps
+        self.hidden_act = hidden_act
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> "PretrainedConfig":
+
+        config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
+
+        # get the vision config dict if we are loading from AltCLIPConfig
+        if config_dict.get("model_type") == "altclip":
+            config_dict = config_dict["vision_config"]
+
+        if "model_type" in config_dict and hasattr(cls, "model_type") and config_dict["model_type"] != cls.model_type:
+            logger.warning(
+                f"You are using a model of type {config_dict['model_type']} to instantiate a model of type "
+                f"{cls.model_type}. This is not supported for all configurations of models and can yield errors."
+            )
+
+        return cls.from_dict(config_dict, **kwargs)
+
+
+class AltCLIPConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`AltCLIPModel`]. It is used to instantiate an
+    AltCLIP 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 AltCLIP
+    [BAAI/AltCLIP](https://huggingface.co/BAAI/AltCLIP) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        text_config (`dict`, *optional*):
+            Dictionary of configuration options used to initialize [`AltCLIPTextConfig`].
+        vision_config (`dict`, *optional*):
+            Dictionary of configuration options used to initialize [`AltCLIPVisionConfig`].
+        projection_dim (`int`, *optional*, defaults to 512):
+            Dimentionality of text and vision projection layers.
+        logit_scale_init_value (`float`, *optional*, defaults to 2.6592):
+            The inital value of the *logit_scale* paramter. Default is used as per the original CLIP implementation.
+        kwargs (*optional*):
+            Dictionary of keyword arguments.
+
+    Example:
+
+    ```python
+    >>> from transformers import AltCLIPConfig, AltCLIPModel
+
+    >>> # Initializing a AltCLIPConfig with BAAI/AltCLIP style configuration
+    >>> configuration = AltCLIPConfig()
+
+    >>> # Initializing a AltCLIPModel (with random weights) from the BAAI/AltCLIP style configuration
+    >>> model = AltCLIPModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+
+    >>> # We can also initialize a AltCLIPConfig from a AltCLIPTextConfig and a AltCLIPVisionConfig
+
+    >>> # Initializing a AltCLIPText and AltCLIPVision configuration
+    >>> config_text = AltCLIPTextConfig()
+    >>> config_vision = AltCLIPVisionConfig()
+
+    >>> config = AltCLIPConfig.from_text_vision_configs(config_text, config_vision)
+    ```"""
+
+    model_type = "altclip"
+    is_composition = True
+
+    def __init__(
+        self, text_config=None, vision_config=None, projection_dim=768, logit_scale_init_value=2.6592, **kwargs
+    ):
+        super().__init__(**kwargs)
+
+        # If `_config_dict` exist, we use them for the backward compatibility.
+        text_config_dict = kwargs.pop("text_config_dict", None)
+        vision_config_dict = kwargs.pop("vision_config_dict", None)
+        if text_config_dict is not None:
+            text_config = text_config_dict
+        if vision_config_dict is not None:
+            vision_config = vision_config_dict
+
+        if text_config is None:
+            text_config = {}
+            logger.info("text_config is None. Initializing the AltCLIPTextConfig with default values.")
+
+        if vision_config is None:
+            vision_config = {}
+            logger.info("vision_config is None. initializing the AltCLIPVisionConfig with default values.")
+
+        self.text_config = AltCLIPTextConfig(**text_config)
+        self.vision_config = AltCLIPVisionConfig(**vision_config)
+
+        self.projection_dim = projection_dim
+        self.logit_scale_init_value = logit_scale_init_value
+        self.initializer_factor = 1.0
+
+    @classmethod
+    def from_text_vision_configs(cls, text_config: AltCLIPTextConfig, vision_config: AltCLIPVisionConfig, **kwargs):
+        r"""
+        Instantiate a [`AltCLIPConfig`] (or a derived class) from altclip text model configuration and altclip vision
+        model configuration.
+
+        Returns:
+            [`AltCLIPConfig`]: An instance of a configuration object
+        """
+
+        return cls(text_config=text_config.to_dict(), vision_config=vision_config.to_dict(), **kwargs)
+
+    def to_dict(self):
+        """
+        Serializes this instance to a Python dictionary. Override the default [`~PretrainedConfig.to_dict`].
+
+        Returns:
+            `Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
+        """
+        output = copy.deepcopy(self.__dict__)
+        output["text_config"] = self.text_config.to_dict()
+        output["vision_config"] = self.vision_config.to_dict()
+        output["model_type"] = self.__class__.model_type
+        return output
--- a/src/transformers/models/altclip/modeling_altclip.py
+++ b/src/transformers/models/altclip/modeling_altclip.py
--- a/src/transformers/models/altclip/processing_altclip.py
+++ b/src/transformers/models/altclip/processing_altclip.py
@@ -0,0 +1,130 @@
+# coding=utf-8
+# Copyright 2022 WenXiang ZhongzhiCheng LedellWu LiuGuang BoWenZhang 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.
+"""
+Image/Text processor class for AltCLIP
+"""
+import warnings
+
+from ...processing_utils import ProcessorMixin
+from ...tokenization_utils_base import BatchEncoding
+
+
+class AltCLIPProcessor(ProcessorMixin):
+    r"""
+    Constructs a AltCLIP processor which wraps a CLIP image processor and a XLM-Roberta tokenizer into a single
+    processor.
+
+    [`AltCLIPProcessor`] offers all the functionalities of [`CLIPImageProcessor`] and [`XLMRobertaTokenizerFast`]. See
+    the [`~AltCLIPProcessor.__call__`] and [`~AltCLIPProcessor.decode`] for more information.
+
+    Args:
+        image_processor ([`CLIPImageProcessor`]):
+            The image processor is a required input.
+        tokenizer ([`XLMRobertaTokenizerFast`]):
+            The tokenizer is a required input.
+    """
+    attributes = ["image_processor", "tokenizer"]
+    image_processor_class = "CLIPImageProcessor"
+    tokenizer_class = ("XLMRobertaTokenizer", "XLMRobertaTokenizerFast")
+
+    def __init__(self, image_processor=None, tokenizer=None, **kwargs):
+        if "feature_extractor" in kwargs:
+            warnings.warn(
+                "The `feature_extractor` argument is deprecated and will be removed in v5, use `image_processor`"
+                " instead.",
+                FutureWarning,
+            )
+            feature_extractor = kwargs.pop("feature_extractor")
+
+        image_processor = image_processor if image_processor is not None else feature_extractor
+        if image_processor is None:
+            raise ValueError("You need to specify an `image_processor`.")
+        if tokenizer is None:
+            raise ValueError("You need to specify a `tokenizer`.")
+
+        super().__init__(image_processor, tokenizer)
+
+    def __call__(self, text=None, images=None, return_tensors=None, **kwargs):
+        """
+        Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
+        and `kwargs` arguments to XLMRobertaTokenizerFast's [`~XLMRobertaTokenizerFast.__call__`] if `text` is not
+        `None` to encode the text. To prepare the image(s), this method forwards the `images` and `kwrags` arguments to
+        CLIPImageProcessor's [`~CLIPImageProcessor.__call__`] if `images` is not `None`. Please refer to the doctsring
+        of the above two methods for more information.
+
+        Args:
+            text (`str`, `List[str]`, `List[List[str]]`):
+                The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
+                (pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
+                `is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
+            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 [`~utils.TensorType`], *optional*):
+                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:
+            [`BatchEncoding`]: A [`BatchEncoding`] with the following fields:
+
+            - **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`.
+            - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when
+              `return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names` and if `text` is not
+              `None`).
+            - **pixel_values** -- Pixel values to be fed to a model. Returned when `images` is not `None`.
+        """
+
+        if text is None and images is None:
+            raise ValueError("You have to specify either text or images. Both cannot be none.")
+
+        if text is not None:
+            encoding = self.tokenizer(text, return_tensors=return_tensors, **kwargs)
+
+        if images is not None:
+            image_features = self.image_processor(images, return_tensors=return_tensors, **kwargs)
+
+        if text is not None and images is not None:
+            encoding["pixel_values"] = image_features.pixel_values
+            return encoding
+        elif text is not None:
+            return encoding
+        else:
+            return BatchEncoding(data=dict(**image_features), tensor_type=return_tensors)
+
+    def batch_decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to XLMRobertaTokenizerFast's [`~PreTrainedTokenizer.batch_decode`].
+        Please refer to the docstring of this method for more information.
+        """
+        return self.tokenizer.batch_decode(*args, **kwargs)
+
+    def decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to XLMRobertaTokenizerFast's [`~PreTrainedTokenizer.decode`]. Please
+        refer to the docstring of this method for more information.
+        """
+        return self.tokenizer.decode(*args, **kwargs)
+
+    @property
+    def model_input_names(self):
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = self.image_processor.model_input_names
+        return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))
--- a/src/transformers/models/auto/configuration_auto.py
+++ b/src/transformers/models/auto/configuration_auto.py
@@ -30,6 +30,7 @@ CONFIG_MAPPING_NAMES = OrderedDict(
    [
        # Add configs here
        ("albert", "AlbertConfig"),
+        ("altclip", "AltCLIPConfig"),
        ("audio-spectrogram-transformer", "ASTConfig"),
        ("bart", "BartConfig"),
        ("beit", "BeitConfig"),
@@ -191,6 +192,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict(
    [
        # Add archive maps here)
        ("albert", "ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
+        ("altclip", "ALTCLIP_PRETRAINED_CONFIG_ARCHIVE_MAP"),
        ("audio-spectrogram-transformer", "AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP"),
        ("bart", "BART_PRETRAINED_CONFIG_ARCHIVE_MAP"),
        ("beit", "BEIT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
@@ -335,6 +337,7 @@ MODEL_NAMES_MAPPING = OrderedDict(
    [
        # Add full (and cased) model names here
        ("albert", "ALBERT"),
+        ("altclip", "AltCLIP"),
        ("audio-spectrogram-transformer", "Audio Spectrogram Transformer"),
        ("bart", "BART"),
        ("barthez", "BARThez"),
--- a/src/transformers/models/auto/modeling_auto.py
+++ b/src/transformers/models/auto/modeling_auto.py
@@ -29,6 +29,7 @@ MODEL_MAPPING_NAMES = OrderedDict(
    [
        # Base model mapping
        ("albert", "AlbertModel"),
+        ("altclip", "AltCLIPModel"),
        ("audio-spectrogram-transformer", "ASTModel"),
        ("bart", "BartModel"),
        ("beit", "BeitModel"),
@@ -878,6 +879,7 @@ MODEL_FOR_AUDIO_XVECTOR_MAPPING_NAMES = OrderedDict(
 _MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPING_NAMES = OrderedDict(
    [
        # Model for Zero Shot Image Classification mapping
+        ("altclip", "AltCLIPModel"),
        ("blip", "BlipModel"),
        ("chinese_clip", "ChineseCLIPModel"),
        ("clip", "CLIPModel"),
--- a/src/transformers/models/auto/processing_auto.py
+++ b/src/transformers/models/auto/processing_auto.py
@@ -41,6 +41,7 @@ logger = logging.get_logger(__name__)

 PROCESSOR_MAPPING_NAMES = OrderedDict(
    [
+        ("altclip", "AltCLIPProcessor"),
        ("blip", "BLIPProcessor"),
        ("chinese_clip", "ChineseCLIPProcessor"),
        ("clip", "CLIPProcessor"),
--- a/src/transformers/utils/dummy_pt_objects.py
+++ b/src/transformers/utils/dummy_pt_objects.py
@@ -357,6 +357,37 @@ def load_tf_weights_in_albert(*args, **kwargs):
    requires_backends(load_tf_weights_in_albert, ["torch"])


+ALTCLIP_PRETRAINED_MODEL_ARCHIVE_LIST = None
+
+
+class AltCLIPModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class AltCLIPPreTrainedModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class AltCLIPTextModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class AltCLIPVisionModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
 AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_MODEL_ARCHIVE_LIST = None


--- a/src/transformers/utils/fx.py
+++ b/src/transformers/utils/fx.py
@@ -101,6 +101,7 @@ def _generate_supported_model_class_names(


 _REGULAR_SUPPORTED_MODEL_NAMES_AND_TASKS = [
+    "altclip",
    "albert",
    "bart",
    "bert",
@@ -156,6 +157,8 @@ _SPECIAL_SUPPORTED_MODELS = [
    "CLIPTextModelWithProjection",
    "CLIPVisionModel",
    "CLIPVisionModelWithProjection",
+    "AltCLIPTextModel",
+    "AltCLIPVisionModel",
    "GitVisionModel",
    "GPT2DoubleHeadsModel",
    "Speech2Text2Decoder",
--- a/tests/models/altclip/init.py
+++ b/tests/models/altclip/init.py
--- a/tests/models/altclip/test_modeling_altclip.py
+++ b/tests/models/altclip/test_modeling_altclip.py
@@ -0,0 +1,539 @@
+# 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.
+""" Testing suite for the PyTorch AltCLIP model. """
+
+
+import inspect
+import os
+import tempfile
+import unittest
+
+import numpy as np
+
+import requests
+from transformers import AltCLIPConfig, AltCLIPProcessor, AltCLIPTextConfig, AltCLIPVisionConfig
+from transformers.testing_utils import require_torch, require_vision, slow, torch_device
+from transformers.utils import is_torch_available, is_vision_available
+
+from ...test_configuration_common import ConfigTester
+from ...test_modeling_common import (
+    ModelTesterMixin,
+    _config_zero_init,
+    floats_tensor,
+    ids_tensor,
+    random_attention_mask,
+)
+
+
+if is_torch_available():
+    import torch
+    import torch.nn as nn
+
+    from transformers import AltCLIPModel, AltCLIPTextModel, AltCLIPVisionModel
+    from transformers.models.altclip.modeling_altclip import ALTCLIP_PRETRAINED_MODEL_ARCHIVE_LIST
+
+if is_vision_available():
+    from PIL import Image
+
+
+class AltCLIPVisionModelTester:
+    def __init__(
+        self,
+        parent,
+        batch_size=12,
+        image_size=30,
+        patch_size=2,
+        num_channels=3,
+        is_training=True,
+        hidden_size=32,
+        projection_dim=32,
+        num_hidden_layers=5,
+        num_attention_heads=4,
+        intermediate_size=37,
+        dropout=0.1,
+        attention_dropout=0.1,
+        initializer_range=0.02,
+        scope=None,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.image_size = image_size
+        self.patch_size = patch_size
+        self.num_channels = num_channels
+        self.is_training = is_training
+        self.hidden_size = hidden_size
+        self.projection_dim = projection_dim
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.dropout = dropout
+        self.attention_dropout = attention_dropout
+        self.initializer_range = initializer_range
+        self.scope = scope
+
+        # in ViT, the seq length equals the number of patches + 1 (we add 1 for the [CLS] token)
+        num_patches = (image_size // patch_size) ** 2
+        self.seq_length = num_patches + 1
+
+    def prepare_config_and_inputs(self):
+        pixel_values = floats_tensor([self.batch_size, self.num_channels, self.image_size, self.image_size])
+        config = self.get_config()
+
+        return config, pixel_values
+
+    def get_config(self):
+        return AltCLIPVisionConfig(
+            image_size=self.image_size,
+            patch_size=self.patch_size,
+            num_channels=self.num_channels,
+            hidden_size=self.hidden_size,
+            projection_dim=self.projection_dim,
+            num_hidden_layers=self.num_hidden_layers,
+            num_attention_heads=self.num_attention_heads,
+            intermediate_size=self.intermediate_size,
+            dropout=self.dropout,
+            attention_dropout=self.attention_dropout,
+            initializer_range=self.initializer_range,
+        )
+
+    def create_and_check_model(self, config, pixel_values):
+        model = AltCLIPVisionModel(config=config)
+        model.to(torch_device)
+        model.eval()
+        with torch.no_grad():
+            result = model(pixel_values)
+        # expected sequence length = num_patches + 1 (we add 1 for the [CLS] token)
+        image_size = (self.image_size, self.image_size)
+        patch_size = (self.patch_size, self.patch_size)
+        num_patches = (image_size[1] // patch_size[1]) * (image_size[0] // patch_size[0])
+        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, num_patches + 1, self.hidden_size))
+        self.parent.assertEqual(result.pooler_output.shape, (self.batch_size, self.hidden_size))
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        config, pixel_values = config_and_inputs
+        inputs_dict = {"pixel_values": pixel_values}
+        return config, inputs_dict
+
+
+@require_torch
+class AltCLIPVisionModelTest(ModelTesterMixin, unittest.TestCase):
+    """
+    Here we also overwrite some of the tests of test_modeling_common.py, as CLIP does not use input_ids, inputs_embeds,
+    attention_mask and seq_length.
+    """
+
+    all_model_classes = (AltCLIPVisionModel,) if is_torch_available() else ()
+    fx_compatible = False
+    test_pruning = False
+    test_resize_embeddings = False
+    test_head_masking = False
+
+    def setUp(self):
+        self.model_tester = AltCLIPVisionModelTester(self)
+        self.config_tester = ConfigTester(
+            self, config_class=AltCLIPVisionConfig, has_text_modality=False, hidden_size=37
+        )
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    @unittest.skip(reason="CLIP does not use inputs_embeds")
+    def test_inputs_embeds(self):
+        pass
+
+    def test_model_common_attributes(self):
+        config, _ = self.model_tester.prepare_config_and_inputs_for_common()
+
+        for model_class in self.all_model_classes:
+            model = model_class(config)
+            self.assertIsInstance(model.get_input_embeddings(), (nn.Module))
+            x = model.get_output_embeddings()
+            self.assertTrue(x is None or isinstance(x, nn.Linear))
+
+    def test_forward_signature(self):
+        config, _ = self.model_tester.prepare_config_and_inputs_for_common()
+
+        for model_class in self.all_model_classes:
+            model = model_class(config)
+            signature = inspect.signature(model.forward)
+            # signature.parameters is an OrderedDict => so arg_names order is deterministic
+            arg_names = [*signature.parameters.keys()]
+
+            expected_arg_names = ["pixel_values"]
+            self.assertListEqual(arg_names[:1], expected_arg_names)
+
+    def test_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_model(*config_and_inputs)
+
+    def test_training(self):
+        pass
+
+    def test_training_gradient_checkpointing(self):
+        pass
+
+    @unittest.skip(reason="AltCLIPVisionModel has no base class and is not available in MODEL_MAPPING")
+    def test_save_load_fast_init_from_base(self):
+        pass
+
+    @unittest.skip(reason="AltCLIPVisionModel has no base class and is not available in MODEL_MAPPING")
+    def test_save_load_fast_init_to_base(self):
+        pass
+
+    @unittest.skip(reason="AltCLIPVisionModel use the same cv backbone with CLIP model.")
+    def test_model_from_pretrained(self):
+        pass
+
+
+class AltCLIPTextModelTester:
+    def __init__(
+        self,
+        parent,
+        batch_size=12,
+        seq_length=7,
+        is_training=True,
+        use_input_mask=True,
+        use_labels=True,
+        vocab_size=99,
+        hidden_size=32,
+        projection_dim=32,
+        project_dim=32,
+        num_hidden_layers=5,
+        num_attention_heads=4,
+        intermediate_size=37,
+        dropout=0.1,
+        attention_dropout=0.1,
+        max_position_embeddings=512,
+        initializer_range=0.02,
+        scope=None,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.seq_length = seq_length
+        self.is_training = is_training
+        self.use_input_mask = use_input_mask
+        self.use_labels = use_labels
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.projection_dim = projection_dim
+        self.project_dim = project_dim
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.dropout = dropout
+        self.attention_dropout = attention_dropout
+        self.max_position_embeddings = max_position_embeddings
+        self.initializer_range = initializer_range
+        self.scope = scope
+
+    def prepare_config_and_inputs(self):
+        input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
+
+        input_mask = None
+        if self.use_input_mask:
+            input_mask = random_attention_mask([self.batch_size, self.seq_length])
+
+        if input_mask is not None:
+            batch_size, seq_length = input_mask.shape
+            rnd_start_indices = np.random.randint(1, seq_length - 1, size=(batch_size,))
+            for batch_idx, start_index in enumerate(rnd_start_indices):
+                input_mask[batch_idx, :start_index] = 1
+                input_mask[batch_idx, start_index:] = 0
+
+        config = self.get_config()
+
+        return config, input_ids, input_mask
+
+    def get_config(self):
+        return AltCLIPTextConfig(
+            vocab_size=self.vocab_size,
+            hidden_size=self.hidden_size,
+            projection_dim=self.projection_dim,
+            project_dim=self.project_dim,
+            num_hidden_layers=self.num_hidden_layers,
+            num_attention_heads=self.num_attention_heads,
+            intermediate_size=self.intermediate_size,
+            dropout=self.dropout,
+            attention_dropout=self.attention_dropout,
+            max_position_embeddings=self.max_position_embeddings,
+            initializer_range=self.initializer_range,
+            pad_token_id=1,
+        )
+
+    def create_and_check_model(self, config, input_ids, input_mask):
+        model = AltCLIPTextModel(config=config)
+        model.to(torch_device)
+        model.eval()
+        with torch.no_grad():
+            result = model(input_ids, attention_mask=input_mask)
+            result = model(input_ids)
+        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
+        self.parent.assertEqual(result.pooler_output.shape, (self.batch_size, self.projection_dim))
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        config, input_ids, input_mask = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask}
+        return config, inputs_dict
+
+
+@require_torch
+class AltCLIPTextModelTest(ModelTesterMixin, unittest.TestCase):
+
+    all_model_classes = (AltCLIPTextModel,) if is_torch_available() else ()
+    fx_compatible = True
+    test_pruning = False
+    test_head_masking = False
+
+    def setUp(self):
+        self.model_tester = AltCLIPTextModelTester(self)
+        self.config_tester = ConfigTester(self, config_class=AltCLIPTextConfig, hidden_size=37)
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    def test_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_model(*config_and_inputs)
+
+    def test_training(self):
+        pass
+
+    def test_training_gradient_checkpointing(self):
+        pass
+
+    def test_model_outputs_equivalence(self):
+        pass
+
+    @unittest.skip(reason="Result of the model is a dict")
+    def test_hidden_states_output(self):
+        pass
+
+    @unittest.skip(reason="AltCLIP does not use inputs_embeds")
+    def test_inputs_embeds(self):
+        pass
+
+    @unittest.skip(reason="AltCLIPTextModel has no base class and is not available in MODEL_MAPPING")
+    def test_save_load_fast_init_from_base(self):
+        pass
+
+    @unittest.skip(reason="AltCLIPTextModel has no base class and is not available in MODEL_MAPPING")
+    def test_save_load_fast_init_to_base(self):
+        pass
+
+    @slow
+    def test_model_from_pretrained(self):
+        for model_name in ALTCLIP_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
+            model = AltCLIPTextModel.from_pretrained(model_name)
+            self.assertIsNotNone(model)
+
+
+class AltCLIPModelTester:
+    def __init__(self, parent, text_kwargs=None, vision_kwargs=None, is_training=True):
+
+        if text_kwargs is None:
+            text_kwargs = {}
+        if vision_kwargs is None:
+            vision_kwargs = {}
+
+        self.parent = parent
+        self.text_model_tester = AltCLIPTextModelTester(parent, **text_kwargs)
+        self.vision_model_tester = AltCLIPVisionModelTester(parent, **vision_kwargs)
+        self.is_training = is_training
+
+    def prepare_config_and_inputs(self):
+        text_config, input_ids, attention_mask = self.text_model_tester.prepare_config_and_inputs()
+        vision_config, pixel_values = self.vision_model_tester.prepare_config_and_inputs()
+
+        config = self.get_config()
+        return config, input_ids, attention_mask, pixel_values
+
+    def get_config(self):
+        return AltCLIPConfig.from_text_vision_configs(
+            self.text_model_tester.get_config(), self.vision_model_tester.get_config(), projection_dim=64
+        )
+
+    def create_and_check_model(self, config, input_ids, attention_mask, pixel_values):
+        model = AltCLIPModel(config=config)
+        model.to(torch_device)
+        model.eval()
+
+        with torch.no_grad():
+            model(input_ids, pixel_values, attention_mask)
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        config, input_ids, attention_mask, pixel_values = config_and_inputs
+        inputs_dict = {
+            "input_ids": input_ids,
+            "attention_mask": attention_mask,
+            "pixel_values": pixel_values,
+            "return_loss": True,
+        }
+        return config, inputs_dict
+
+
+# We will verify our results on an image of cute cats
+def prepare_img():
+    url = "http://images.cocodataset.org/val2017/000000039769.jpg"
+    im = Image.open(requests.get(url, stream=True).raw)
+    return im
+
+
+@require_torch
+class AltCLIPModelTest(ModelTesterMixin, unittest.TestCase):
+
+    all_model_classes = (AltCLIPModel,) if is_torch_available() else ()
+    fx_compatible = True
+    test_head_masking = False
+    test_pruning = False
+    test_resize_embeddings = False
+    test_attention_outputs = False
+
+    def setUp(self):
+        self.model_tester = AltCLIPModelTester(self)
+
+    def test_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_model(*config_and_inputs)
+
+    @unittest.skip(reason="Hidden_states is tested in individual model tests")
+    def test_hidden_states_output(self):
+        pass
+
+    @unittest.skip(reason="Inputs_embeds is tested in individual model tests")
+    def test_inputs_embeds(self):
+        pass
+
+    @unittest.skip(reason="Retain_grad is tested in individual model tests")
+    def test_retain_grad_hidden_states_attentions(self):
+        pass
+
+    @unittest.skip(reason="CLIPModel does not have input/output embeddings")
+    def test_model_common_attributes(self):
+        pass
+
+    # override as the `logit_scale` parameter initilization is different for AltCLIP
+    def test_initialization(self):
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+        configs_no_init = _config_zero_init(config)
+        for model_class in self.all_model_classes:
+            model = model_class(config=configs_no_init)
+            for name, param in model.named_parameters():
+                if param.requires_grad:
+                    # check if `logit_scale` is initilized as per the original implementation
+                    if name == "logit_scale":
+                        self.assertAlmostEqual(
+                            param.data.item(),
+                            np.log(1 / 0.07),
+                            delta=1e-3,
+                            msg=f"Parameter {name} of model {model_class} seems not properly initialized",
+                        )
+                    else:
+                        self.assertIn(
+                            ((param.data.mean() * 1e9).round() / 1e9).item(),
+                            [0.0, 1.0],
+                            msg=f"Parameter {name} of model {model_class} seems not properly initialized",
+                        )
+
+    def _create_and_check_torchscript(self, config, inputs_dict):
+        if not self.test_torchscript:
+            return
+
+        configs_no_init = _config_zero_init(config)  # To be sure we have no Nan
+        configs_no_init.torchscript = True
+        configs_no_init.return_dict = False
+        for model_class in self.all_model_classes:
+            model = model_class(config=configs_no_init)
+            model.to(torch_device)
+            model.eval()
+
+            try:
+                input_ids = inputs_dict["input_ids"]
+                pixel_values = inputs_dict["pixel_values"]  # CLIP needs pixel_values
+                traced_model = torch.jit.trace(model, (input_ids, pixel_values))
+            except RuntimeError:
+                self.fail("Couldn't trace module.")
+
+            with tempfile.TemporaryDirectory() as tmp_dir_name:
+                pt_file_name = os.path.join(tmp_dir_name, "traced_model.pt")
+
+                try:
+                    torch.jit.save(traced_model, pt_file_name)
+                except Exception:
+                    self.fail("Couldn't save module.")
+
+                try:
+                    loaded_model = torch.jit.load(pt_file_name)
+                except Exception:
+                    self.fail("Couldn't load module.")
+
+            model.to(torch_device)
+            model.eval()
+
+            loaded_model.to(torch_device)
+            loaded_model.eval()
+
+            model_state_dict = model.state_dict()
+            loaded_model_state_dict = loaded_model.state_dict()
+
+            self.assertEqual(set(model_state_dict.keys()), set(loaded_model_state_dict.keys()))
+
+            models_equal = True
+            for layer_name, p1 in model_state_dict.items():
+                p2 = loaded_model_state_dict[layer_name]
+                if p1.data.ne(p2.data).sum() > 0:
+                    models_equal = False
+
+            self.assertTrue(models_equal)
+
+    @slow
+    def test_model_from_pretrained(self):
+        for model_name in ALTCLIP_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
+            model = AltCLIPModel.from_pretrained(model_name)
+            self.assertIsNotNone(model)
+
+
+@require_vision
+@require_torch
+class AltCLIPModelIntegrationTest(unittest.TestCase):
+    @slow
+    def test_inference(self):
+        model_name = "BAAI/AltCLIP"
+        model = AltCLIPModel.from_pretrained(model_name).to(torch_device)
+        processor = AltCLIPProcessor.from_pretrained(model_name)
+
+        image = prepare_img()
+        inputs = processor(text=["一张猫的照片", "一张狗的照片"], images=image, padding=True, return_tensors="pt").to(torch_device)
+
+        # forward pass
+        with torch.no_grad():
+            outputs = model(**inputs)
+
+        # verify the logits
+        self.assertEqual(
+            outputs.logits_per_image.shape,
+            torch.Size((inputs.pixel_values.shape[0], inputs.input_ids.shape[0])),
+        )
+        self.assertEqual(
+            outputs.logits_per_text.shape,
+            torch.Size((inputs.input_ids.shape[0], inputs.pixel_values.shape[0])),
+        )
+
+        probs = outputs.logits_per_image.softmax(dim=1)
+        expected_probs = torch.tensor([[9.9942e-01, 5.7805e-04]], device=torch_device)
+
+        self.assertTrue(torch.allclose(probs, expected_probs, atol=5e-3))
--- a/utils/check_repo.py
+++ b/utils/check_repo.py
@@ -35,6 +35,7 @@ PATH_TO_DOC = "docs/source/en"

 # Update this list with models that are supposed to be private.
 PRIVATE_MODELS = [
+    "AltRobertaModel",
    "DPRSpanPredictor",
    "LongT5Stack",
    "RealmBertModel",
@@ -121,6 +122,7 @@ IGNORE_NON_TESTED = PRIVATE_MODELS.copy() + [
    "FlaxBertForCausalLM",  # Building part of bigger (tested) model. Tested implicitly through FlaxRobertaForCausalLM.
    "OPTDecoderWrapper",
    "TFSegformerDecodeHead",  # Not a regular model.
+    "AltRobertaModel",  # Building part of bigger (tested) model.
    "BlipTextLMHeadModel",  # No need to test it as it is tested by BlipTextVision models
 ]

@@ -251,6 +253,9 @@ IGNORE_NON_AUTO_CONFIGURED = PRIVATE_MODELS.copy() + [
    "TFHubertForCTC",
    "XCLIPVisionModel",
    "XCLIPTextModel",
+    "AltCLIPTextModel",
+    "AltCLIPVisionModel",
+    "AltRobertaModel",
 ]

 # Update this list for models that have multiple model types for the same
@@ -673,6 +678,7 @@ UNDOCUMENTED_OBJECTS = [
    "logger",  # Internal logger
    "logging",  # External module
    "requires_backends",  # Internal function
+    "AltRobertaModel",  # Internal module
 ]

 # This list should be empty. Objects in it should get their own doc page.