Add support for conditional detr (#18948)

* added conditional_detr files

* checked copies

* checked copies

* fixed style and copies

* fixed style and copies

* fixed hub

* fixed style

* Update README.md
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/_toctree.yml
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/index.mdx
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/convert_conditional_detr_original_pytorch_checkpoint_to_pytorch.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/model_doc/conditional_detr.mdx
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* fixed some doc issue

* changed prefix to ConditionalDetr

* fixed docs

* Update README_ko.md

* added spatial_model_name

* fixed fix-copies

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added some copied from

* added some copied from

* added some copied from

* added some copied from

* fixed use_pretrained issue

* changed post-process

* added conditional_detr files

* checked copies

* checked copies

* fixed style and copies

* fixed style and copies

* fixed hub

* fixed style

* Update README.md
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/_toctree.yml
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/index.mdx
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/convert_conditional_detr_original_pytorch_checkpoint_to_pytorch.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* fixed some doc issue

* Update docs/source/en/model_doc/conditional_detr.mdx
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* changed prefix to ConditionalDetr

* fixed docs

* Update README_ko.md

* added spatial_model_name

* fixed fix-copies

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added some copied from

* added some copied from

* added some copied from

* added some copied from

* fixed use_pretrained issue

* changed post-process

* fix style quality and copies

* fix style quality and copies

* fix style quality and copies

* fix style quality and copies

* add more fix-copies

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* fixed some variable names & added more fix-copies

* fixed some variable names & added more fix-copies

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added more copied from

* fixed quality

* changed pretrained config

* added more copied-from and fixed the issue in feature_extraction_auto

* added conditional_detr files

* checked copies

* checked copies

* fixed style and copies

* fixed style and copies

* fixed hub

* fixed style

* Update README.md
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/_toctree.yml
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/index.mdx
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/convert_conditional_detr_original_pytorch_checkpoint_to_pytorch.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* fixed some doc issue

* Update docs/source/en/model_doc/conditional_detr.mdx
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* changed prefix to ConditionalDetr

* fixed docs

* Update README_ko.md

* added spatial_model_name

* fixed fix-copies

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added some copied from

* added some copied from

* added some copied from

* added some copied from

* fixed use_pretrained issue

* changed post-process

* added conditional_detr files

* checked copies

* fixed style and copies

* fixed some doc issue

* changed prefix to ConditionalDetr

* fixed docs

* added spatial_model_name

* fixed fix-copies

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added some copied from

* added some copied from

* added some copied from

* added some copied from

* fix style quality and copies

* fix style quality and copies

* fix style quality and copies

* add more fix-copies

* fixed some variable names & added more fix-copies

* fixed some variable names & added more fix-copies

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added more copied from

* fixed quality

* changed pretrained config

* added more copied-from and fixed the issue in feature_extraction_auto

* fixed style

* added conditional_detr files

* checked copies

* checked copies

* fixed style and copies

* fixed style and copies

* fixed hub

* fixed style

* Update README.md
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/_toctree.yml
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update docs/source/en/index.mdx
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/convert_conditional_detr_original_pytorch_checkpoint_to_pytorch.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* fixed some doc issue

* Update docs/source/en/model_doc/conditional_detr.mdx
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* changed prefix to ConditionalDetr

* fixed docs

* Update README_ko.md

* added spatial_model_name

* fixed fix-copies

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added some copied from

* added some copied from

* added some copied from

* added some copied from

* fixed use_pretrained issue

* changed post-process

* added conditional_detr files

* checked copies

* fixed style and copies

* fixed some doc issue

* changed prefix to ConditionalDetr

* fixed docs

* added spatial_model_name

* fixed fix-copies

* Update src/transformers/models/conditional_detr/modeling_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added some copied from

* added some copied from

* added some copied from

* added some copied from

* fix style quality and copies

* fix style quality and copies

* fix style quality and copies

* add more fix-copies

* fixed some variable names & added more fix-copies

* fixed some variable names & added more fix-copies

* Update src/transformers/models/conditional_detr/feature_extraction_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/conditional_detr/configuration_conditional_detr.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* added more copied from

* fixed quality

* changed pretrained config

* added more copied-from and fixed the issue in feature_extraction_auto

* rebased
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>
Co-authored-by: Depu Meng <depumeng@Depus-MacBook-Pro.local>

README.md

@@ -278,6 +278,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
 1. **[CLIP](https://huggingface.co/docs/transformers/model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
 1. **[CodeGen](https://huggingface.co/docs/transformers/model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[Conditional DETR](https://huggingface.co/docs/transformers/main/model_doc/conditional_detr)** (from Microsoft Research Asia) released with the paper [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang.
 1. **[ConvBERT](https://huggingface.co/docs/transformers/model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
 1. **[ConvNeXT](https://huggingface.co/docs/transformers/model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
 1. **[CPM](https://huggingface.co/docs/transformers/model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.

README_ko.md

@@ -228,6 +228,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
 1. **[CLIP](https://huggingface.co/docs/transformers/model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
 1. **[CodeGen](https://huggingface.co/docs/transformers/model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[Conditional DETR](https://huggingface.co/docs/transformers/main/model_doc/conditional_detr)** (from Microsoft Research Asia) released with the paper [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang.
 1. **[ConvBERT](https://huggingface.co/docs/transformers/model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
 1. **[ConvNeXT](https://huggingface.co/docs/transformers/model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
 1. **[CPM](https://huggingface.co/docs/transformers/model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.

README_zh-hans.md

@@ -252,6 +252,7 @@ conda install -c huggingface transformers
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (来自 Google Research) 伴随论文 [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) 由 Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting 发布。
 1. **[CLIP](https://huggingface.co/docs/transformers/model_doc/clip)** (来自 OpenAI) 伴随论文 [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) 由 Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever 发布。
 1. **[CodeGen](https://huggingface.co/docs/transformers/model_doc/codegen)** (来自 Salesforce) 伴随论文 [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) 由 Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong 发布。
+1. **[Conditional DETR](https://huggingface.co/docs/transformers/main/model_doc/conditional_detr)** (来自 Microsoft Research Asia) 伴随论文 [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) 由 Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang 发布。
 1. **[ConvBERT](https://huggingface.co/docs/transformers/model_doc/convbert)** (来自 YituTech) 伴随论文 [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) 由 Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan 发布。
 1. **[ConvNeXT](https://huggingface.co/docs/transformers/model_doc/convnext)** (来自 Facebook AI) 伴随论文 [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) 由 Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie 发布。
 1. **[CPM](https://huggingface.co/docs/transformers/model_doc/cpm)** (来自 Tsinghua University) 伴随论文 [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) 由 Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun 发布。

README_zh-hant.md

@@ -264,6 +264,7 @@ conda install -c huggingface transformers
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
 1. **[CLIP](https://huggingface.co/docs/transformers/model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
 1. **[CodeGen](https://huggingface.co/docs/transformers/model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[Conditional DETR](https://huggingface.co/docs/transformers/main/model_doc/conditional_detr)** (from Microsoft Research Asia) released with the paper [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang.
 1. **[ConvBERT](https://huggingface.co/docs/transformers/model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
 1. **[ConvNeXT](https://huggingface.co/docs/transformers/model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
 1. **[CPM](https://huggingface.co/docs/transformers/model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.

docs/source/en/_toctree.yml

@@ -362,6 +362,8 @@
       sections:
       - local: model_doc/beit
         title: BEiT
+      - local: model_doc/conditional_detr
+        title: Conditional DETR
       - local: model_doc/convnext
         title: ConvNeXT
       - local: model_doc/cvt

docs/source/en/index.mdx

@@ -68,6 +68,7 @@ The documentation is organized into five sections:
 1. **[CANINE](model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
 1. **[CLIP](model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
 1. **[CodeGen](model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[Conditional DETR](model_doc/conditional_detr)** (from Microsoft Research Asia) released with the paper [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang.
 1. **[ConvBERT](model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
 1. **[ConvNeXT](model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
 1. **[CPM](model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.
@@ -215,6 +216,7 @@ Flax), PyTorch, and/or TensorFlow.
 |           CANINE            |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            CLIP             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |           CodeGen           |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|      Conditional DETR       |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          ConvBERT           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |          ConvNeXT           |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
 |            CTRL             |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |

docs/source/en/model_doc/conditional_detr.mdx

+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Conditional DETR
+
+## Overview
+
+The Conditional DETR model was proposed in [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang. Conditional DETR presents a conditional cross-attention mechanism for fast DETR training. Conditional DETR converges 6.7× to 10× faster than DETR.
+
+The abstract from the paper is the following:
+
+*The recently-developed DETR approach applies the transformer encoder and decoder architecture to object detection and achieves promising performance. In this paper, we handle the critical issue, slow training convergence, and present a conditional cross-attention mechanism for fast DETR training. Our approach is motivated by that the cross-attention in DETR relies highly on the content embeddings for localizing the four extremities and predicting the box, which increases the need for high-quality content embeddings and thus the training difficulty. Our approach, named conditional DETR, learns a conditional spatial query from the decoder embedding for decoder multi-head cross-attention. The benefit is that through the conditional spatial query, each cross-attention head is able to attend to a band containing a distinct region, e.g., one object extremity or a region inside the object box. This narrows down the spatial range for localizing the distinct regions for object classification and box regression, thus relaxing the dependence on the content embeddings and easing the training. Empirical results show that conditional DETR converges 6.7× faster for the backbones R50 and R101 and 10× faster for stronger backbones DC5-R50 and DC5-R101. Code is available at https://github.com/Atten4Vis/ConditionalDETR.*
+
+
+This model was contributed by [DepuMeng](https://huggingface.co/DepuMeng). The original code can be found [here](https://github.com/Atten4Vis/ConditionalDETR).
+
+
+## ConditionalDetrConfig
+
+[[autodoc]] ConditionalDetrConfig
+
+## ConditionalDetrFeatureExtractor
+
+[[autodoc]] ConditionalDetrFeatureExtractor
+    - __call__
+    - pad_and_create_pixel_mask
+    - post_process
+    - post_process_segmentation
+    - post_process_panoptic
+
+## ConditionalDetrModel
+
+[[autodoc]] ConditionalDetrModel
+    - forward
+
+## ConditionalDetrForObjectDetection
+
+[[autodoc]] ConditionalDetrForObjectDetection
+    - forward
+
+## ConditionalDetrForSegmentation
+
+[[autodoc]] ConditionalDetrForSegmentation
+    - forward
\ No newline at end of file

docs/source/en/serialization.mdx

@@ -57,6 +57,7 @@ Ready-made configurations include the following architectures:
 - CamemBERT
 - CLIP
 - CodeGen
+- Conditional DETR
 - ConvBERT
 - ConvNeXT
 - Data2VecText

src/transformers/__init__.py

@@ -172,6 +172,7 @@ _import_structure = {
         "CLIPVisionConfig",
     ],
     "models.codegen": ["CODEGEN_PRETRAINED_CONFIG_ARCHIVE_MAP", "CodeGenConfig", "CodeGenTokenizer"],
+    "models.conditional_detr": ["CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP", "ConditionalDetrConfig"],
     "models.convbert": ["CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "ConvBertConfig", "ConvBertTokenizer"],
     "models.convnext": ["CONVNEXT_PRETRAINED_CONFIG_ARCHIVE_MAP", "ConvNextConfig"],
     "models.cpm": [],
@@ -660,6 +661,7 @@ else:
     _import_structure["models.convnext"].append("ConvNextFeatureExtractor")
     _import_structure["models.deit"].append("DeiTFeatureExtractor")
     _import_structure["models.detr"].append("DetrFeatureExtractor")
+    _import_structure["models.conditional_detr"].append("ConditionalDetrFeatureExtractor")
     _import_structure["models.donut"].append("DonutFeatureExtractor")
     _import_structure["models.dpt"].append("DPTFeatureExtractor")
     _import_structure["models.flava"].extend(["FlavaFeatureExtractor", "FlavaProcessor"])
@@ -708,6 +710,15 @@ else:
             "DetrPreTrainedModel",
         ]
     )
+    _import_structure["models.conditional_detr"].extend(
+        [
+            "CONDITIONAL_DETR_PRETRAINED_MODEL_ARCHIVE_LIST",
+            "ConditionalDetrForObjectDetection",
+            "ConditionalDetrForSegmentation",
+            "ConditionalDetrModel",
+            "ConditionalDetrPreTrainedModel",
+        ]
+    )
 
 try:
     if not is_scatter_available():
@@ -3075,6 +3086,7 @@ if TYPE_CHECKING:
         CLIPVisionConfig,
     )
     from .models.codegen import CODEGEN_PRETRAINED_CONFIG_ARCHIVE_MAP, CodeGenConfig, CodeGenTokenizer
+    from .models.conditional_detr import CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP, ConditionalDetrConfig
     from .models.convbert import CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, ConvBertConfig, ConvBertTokenizer
     from .models.convnext import CONVNEXT_PRETRAINED_CONFIG_ARCHIVE_MAP, ConvNextConfig
     from .models.ctrl import CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP, CTRLConfig, CTRLTokenizer
@@ -3498,6 +3510,7 @@ if TYPE_CHECKING:
         from .image_utils import ImageFeatureExtractionMixin
         from .models.beit import BeitFeatureExtractor
         from .models.clip import CLIPFeatureExtractor
+        from .models.conditional_detr import ConditionalDetrFeatureExtractor
         from .models.convnext import ConvNextFeatureExtractor
         from .models.deit import DeiTFeatureExtractor
         from .models.detr import DetrFeatureExtractor
@@ -3527,6 +3540,13 @@ if TYPE_CHECKING:
     except OptionalDependencyNotAvailable:
         from .utils.dummy_timm_and_vision_objects import *
     else:
+        from .models.conditional_detr import (
+            CONDITIONAL_DETR_PRETRAINED_MODEL_ARCHIVE_LIST,
+            ConditionalDetrForObjectDetection,
+            ConditionalDetrForSegmentation,
+            ConditionalDetrModel,
+            ConditionalDetrPreTrainedModel,
+        )
         from .models.deformable_detr import (
             DEFORMABLE_DETR_PRETRAINED_MODEL_ARCHIVE_LIST,
             DeformableDetrForObjectDetection,

src/transformers/models/__init__.py

@@ -38,6 +38,7 @@ from . import (
     canine,
     clip,
     codegen,
+    conditional_detr,
     convbert,
     convnext,
     cpm,

src/transformers/models/auto/configuration_auto.py

@@ -43,6 +43,7 @@ CONFIG_MAPPING_NAMES = OrderedDict(
         ("canine", "CanineConfig"),
         ("clip", "CLIPConfig"),
         ("codegen", "CodeGenConfig"),
+        ("conditional_detr", "ConditionalDetrConfig"),
         ("convbert", "ConvBertConfig"),
         ("convnext", "ConvNextConfig"),
         ("ctrl", "CTRLConfig"),
@@ -175,6 +176,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict(
         ("canine", "CANINE_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("clip", "CLIP_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("codegen", "CODEGEN_PRETRAINED_CONFIG_ARCHIVE_MAP"),
+        ("conditional_detr", "CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("convbert", "CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("convnext", "CONVNEXT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("ctrl", "CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP"),
@@ -300,6 +302,7 @@ MODEL_NAMES_MAPPING = OrderedDict(
         ("canine", "CANINE"),
         ("clip", "CLIP"),
         ("codegen", "CodeGen"),
+        ("conditional_detr", "Conditional DETR"),
         ("convbert", "ConvBERT"),
         ("convnext", "ConvNeXT"),
         ("cpm", "CPM"),

src/transformers/models/auto/feature_extraction_auto.py

@@ -39,6 +39,7 @@ FEATURE_EXTRACTOR_MAPPING_NAMES = OrderedDict(
     [
         ("beit", "BeitFeatureExtractor"),
         ("clip", "CLIPFeatureExtractor"),
+        ("conditional_detr", "ConditionalDetrFeatureExtractor"),
         ("convnext", "ConvNextFeatureExtractor"),
         ("cvt", "ConvNextFeatureExtractor"),
         ("data2vec-audio", "Wav2Vec2FeatureExtractor"),
@@ -46,7 +47,6 @@ FEATURE_EXTRACTOR_MAPPING_NAMES = OrderedDict(
         ("deformable_detr", "DetrFeatureExtractor"),
         ("deit", "DeiTFeatureExtractor"),
         ("detr", "DetrFeatureExtractor"),
-        ("detr", "DetrFeatureExtractor"),
         ("donut", "DonutFeatureExtractor"),
         ("dpt", "DPTFeatureExtractor"),
         ("flava", "FlavaFeatureExtractor"),

src/transformers/models/auto/modeling_auto.py

@@ -42,6 +42,7 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("canine", "CanineModel"),
         ("clip", "CLIPModel"),
         ("codegen", "CodeGenModel"),
+        ("conditional_detr", "ConditionalDetrModel"),
         ("convbert", "ConvBertModel"),
         ("convnext", "ConvNextModel"),
         ("ctrl", "CTRLModel"),
@@ -455,6 +456,7 @@ MODEL_FOR_MASKED_LM_MAPPING_NAMES = OrderedDict(
 MODEL_FOR_OBJECT_DETECTION_MAPPING_NAMES = OrderedDict(
     [
         # Model for Object Detection mapping
+        ("conditional_detr", "ConditionalDetrForObjectDetection"),
         ("deformable_detr", "DeformableDetrForObjectDetection"),
         ("detr", "DetrForObjectDetection"),
         ("yolos", "YolosForObjectDetection"),

src/transformers/models/conditional_detr/__init__.py

+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+# Copyright 2022 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import TYPE_CHECKING
+
+from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_timm_available, is_vision_available
+
+
+_import_structure = {
+    "configuration_conditional_detr": [
+        "CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP",
+        "ConditionalDetrConfig",
+        "ConditionalDetrOnnxConfig",
+    ]
+}
+
+try:
+    if not is_vision_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["feature_extraction_conditional_detr"] = ["ConditionalDetrFeatureExtractor"]
+
+try:
+    if not is_timm_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_conditional_detr"] = [
+        "CONDITIONAL_DETR_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "ConditionalDetrForObjectDetection",
+        "ConditionalDetrForSegmentation",
+        "ConditionalDetrModel",
+        "ConditionalDetrPreTrainedModel",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_conditional_detr import (
+        CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        ConditionalDetrConfig,
+        ConditionalDetrOnnxConfig,
+    )
+
+    try:
+        if not is_vision_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .feature_extraction_conditional_detr import ConditionalDetrFeatureExtractor
+
+    try:
+        if not is_timm_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_conditional_detr import (
+            CONDITIONAL_DETR_PRETRAINED_MODEL_ARCHIVE_LIST,
+            ConditionalDetrForObjectDetection,
+            ConditionalDetrForSegmentation,
+            ConditionalDetrModel,
+            ConditionalDetrPreTrainedModel,
+        )
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)

src/transformers/models/conditional_detr/configuration_conditional_detr.py

+# coding=utf-8
+# Copyright 2022 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" Conditional DETR model configuration"""
+
+from collections import OrderedDict
+from typing import Mapping
+
+from packaging import version
+
+from ...configuration_utils import PretrainedConfig
+from ...onnx import OnnxConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "microsoft/conditional-detr-resnet-50": (
+        "https://huggingface.co/microsoft/conditional-detr-resnet-50/resolve/main/config.json"
+    ),
+}
+
+
+class ConditionalDetrConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`ConditionalDetrModel`]. It is used to instantiate
+    a Conditional DETR 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 Conditional DETR
+    [microsoft/conditional-detr-resnet-50](https://huggingface.co/microsoft/conditional-detr-resnet-50) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        num_channels (`int`, *optional*, defaults to 3):
+            The number of input channels.
+        num_queries (`int`, *optional*, defaults to 100):
+            Number of object queries, i.e. detection slots. This is the maximal number of objects
+            [`ConditionalDetrModel`] can detect in a single image. For COCO, we recommend 100 queries.
+        d_model (`int`, *optional*, defaults to 256):
+            Dimension of the layers.
+        encoder_layers (`int`, *optional*, defaults to 6):
+            Number of encoder layers.
+        decoder_layers (`int`, *optional*, defaults to 6):
+            Number of decoder layers.
+        encoder_attention_heads (`int`, *optional*, defaults to 8):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        decoder_attention_heads (`int`, *optional*, defaults to 8):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        decoder_ffn_dim (`int`, *optional*, defaults to 2048):
+            Dimension of the "intermediate" (often named feed-forward) layer in decoder.
+        encoder_ffn_dim (`int`, *optional*, defaults to 2048):
+            Dimension of the "intermediate" (often named feed-forward) layer in decoder.
+        activation_function (`str` or `function`, *optional*, defaults to `"relu"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"silu"` and `"gelu_new"` are supported.
+        dropout (`float`, *optional*, defaults to 0.1):
+            The dropout probability 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.
+        activation_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for activations inside the fully connected layer.
+        init_std (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        init_xavier_std (`float`, *optional*, defaults to 1):
+            The scaling factor used for the Xavier initialization gain in the HM Attention map module.
+        encoder_layerdrop (`float`, *optional*, defaults to 0.0):
+            The LayerDrop probability for the encoder. See the [LayerDrop paper](see https://arxiv.org/abs/1909.11556)
+            for more details.
+        decoder_layerdrop (`float`, *optional*, defaults to 0.0):
+            The LayerDrop probability for the decoder. See the [LayerDrop paper](see https://arxiv.org/abs/1909.11556)
+            for more details.
+        auxiliary_loss (`bool`, *optional*, defaults to `False`):
+            Whether auxiliary decoding losses (loss at each decoder layer) are to be used.
+        position_embedding_type (`str`, *optional*, defaults to `"sine"`):
+            Type of position embeddings to be used on top of the image features. One of `"sine"` or `"learned"`.
+        backbone (`str`, *optional*, defaults to `"resnet50"`):
+            Name of convolutional backbone to use. Supports any convolutional backbone from the timm package. For a
+            list of all available models, see [this
+            page](https://rwightman.github.io/pytorch-image-models/#load-a-pretrained-model).
+        use_pretrained_backbone (`bool`, *optional*, defaults to `True`):
+            Whether to use pretrained weights for the backbone.
+        dilation (`bool`, *optional*, defaults to `False`):
+            Whether to replace stride with dilation in the last convolutional block (DC5).
+        class_cost (`float`, *optional*, defaults to 1):
+            Relative weight of the classification error in the Hungarian matching cost.
+        bbox_cost (`float`, *optional*, defaults to 5):
+            Relative weight of the L1 error of the bounding box coordinates in the Hungarian matching cost.
+        giou_cost (`float`, *optional*, defaults to 2):
+            Relative weight of the generalized IoU loss of the bounding box in the Hungarian matching cost.
+        mask_loss_coefficient (`float`, *optional*, defaults to 1):
+            Relative weight of the Focal loss in the panoptic segmentation loss.
+        dice_loss_coefficient (`float`, *optional*, defaults to 1):
+            Relative weight of the DICE/F-1 loss in the panoptic segmentation loss.
+        bbox_loss_coefficient (`float`, *optional*, defaults to 5):
+            Relative weight of the L1 bounding box loss in the object detection loss.
+        giou_loss_coefficient (`float`, *optional*, defaults to 2):
+            Relative weight of the generalized IoU loss in the object detection loss.
+        eos_coefficient (`float`, *optional*, defaults to 0.1):
+            Relative classification weight of the 'no-object' class in the object detection loss.
+
+    Examples:
+
+    ```python
+    >>> from transformers import ConditionalDetrModel, ConditionalDetrConfig
+
+    >>> # Initializing a Conditional DETR microsoft/conditional-detr-resnet-50 style configuration
+    >>> configuration = ConditionalDetrConfig()
+
+    >>> # Initializing a model from the microsoft/conditional-detr-resnet-50 style configuration
+    >>> model = ConditionalDetrModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+    model_type = "conditional_detr"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    attribute_map = {
+        "hidden_size": "d_model",
+        "num_attention_heads": "encoder_attention_heads",
+    }
+
+    def __init__(
+        self,
+        num_channels=3,
+        num_queries=300,
+        max_position_embeddings=1024,
+        encoder_layers=6,
+        encoder_ffn_dim=2048,
+        encoder_attention_heads=8,
+        decoder_layers=6,
+        decoder_ffn_dim=2048,
+        decoder_attention_heads=8,
+        encoder_layerdrop=0.0,
+        decoder_layerdrop=0.0,
+        is_encoder_decoder=True,
+        activation_function="relu",
+        d_model=256,
+        dropout=0.1,
+        attention_dropout=0.0,
+        activation_dropout=0.0,
+        init_std=0.02,
+        init_xavier_std=1.0,
+        classifier_dropout=0.0,
+        scale_embedding=False,
+        auxiliary_loss=False,
+        position_embedding_type="sine",
+        backbone="resnet50",
+        use_pretrained_backbone=True,
+        dilation=False,
+        class_cost=2,
+        bbox_cost=5,
+        giou_cost=2,
+        mask_loss_coefficient=1,
+        dice_loss_coefficient=1,
+        cls_loss_coefficient=2,
+        bbox_loss_coefficient=5,
+        giou_loss_coefficient=2,
+        focal_alpha=0.25,
+        **kwargs
+    ):
+        self.num_channels = num_channels
+        self.num_queries = num_queries
+        self.max_position_embeddings = max_position_embeddings
+        self.d_model = d_model
+        self.encoder_ffn_dim = encoder_ffn_dim
+        self.encoder_layers = encoder_layers
+        self.encoder_attention_heads = encoder_attention_heads
+        self.decoder_ffn_dim = decoder_ffn_dim
+        self.decoder_layers = decoder_layers
+        self.decoder_attention_heads = decoder_attention_heads
+        self.dropout = dropout
+        self.attention_dropout = attention_dropout
+        self.activation_dropout = activation_dropout
+        self.activation_function = activation_function
+        self.init_std = init_std
+        self.init_xavier_std = init_xavier_std
+        self.encoder_layerdrop = encoder_layerdrop
+        self.decoder_layerdrop = decoder_layerdrop
+        self.num_hidden_layers = encoder_layers
+        self.scale_embedding = scale_embedding  # scale factor will be sqrt(d_model) if True
+        self.auxiliary_loss = auxiliary_loss
+        self.position_embedding_type = position_embedding_type
+        self.backbone = backbone
+        self.use_pretrained_backbone = use_pretrained_backbone
+        self.dilation = dilation
+        # Hungarian matcher
+        self.class_cost = class_cost
+        self.bbox_cost = bbox_cost
+        self.giou_cost = giou_cost
+        # Loss coefficients
+        self.mask_loss_coefficient = mask_loss_coefficient
+        self.dice_loss_coefficient = dice_loss_coefficient
+        self.cls_loss_coefficient = cls_loss_coefficient
+        self.bbox_loss_coefficient = bbox_loss_coefficient
+        self.giou_loss_coefficient = giou_loss_coefficient
+        self.focal_alpha = focal_alpha
+        super().__init__(is_encoder_decoder=is_encoder_decoder, **kwargs)
+
+    @property
+    def num_attention_heads(self) -> int:
+        return self.encoder_attention_heads
+
+    @property
+    def hidden_size(self) -> int:
+        return self.d_model
+
+
+class ConditionalDetrOnnxConfig(OnnxConfig):
+
+    torch_onnx_minimum_version = version.parse("1.11")
+
+    @property
+    def inputs(self) -> Mapping[str, Mapping[int, str]]:
+        return OrderedDict(
+            [
+                ("pixel_values", {0: "batch", 1: "num_channels", 2: "height", 3: "width"}),
+                ("pixel_mask", {0: "batch"}),
+            ]
+        )
+
+    @property
+    def atol_for_validation(self) -> float:
+        return 1e-5
+
+    @property
+    def default_onnx_opset(self) -> int:
+        return 12

src/transformers/models/conditional_detr/convert_conditional_detr_original_pytorch_checkpoint_to_pytorch.py

+# coding=utf-8
+# Copyright 2022 The HuggingFace Inc. team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Convert Conditional DETR checkpoints."""
+
+
+import argparse
+import json
+from collections import OrderedDict
+from pathlib import Path
+
+import torch
+from PIL import Image
+
+import requests
+from huggingface_hub import hf_hub_download
+from transformers import (
+    ConditionalDetrConfig,
+    ConditionalDetrFeatureExtractor,
+    ConditionalDetrForObjectDetection,
+    ConditionalDetrForSegmentation,
+)
+from transformers.utils import logging
+
+
+logging.set_verbosity_info()
+logger = logging.get_logger(__name__)
+
+# here we list all keys to be renamed (original name on the left, our name on the right)
+rename_keys = []
+for i in range(6):
+    # encoder layers: output projection, 2 feedforward neural networks and 2 layernorms
+    rename_keys.append(
+        (f"transformer.encoder.layers.{i}.self_attn.out_proj.weight", f"encoder.layers.{i}.self_attn.out_proj.weight")
+    )
+    rename_keys.append(
+        (f"transformer.encoder.layers.{i}.self_attn.out_proj.bias", f"encoder.layers.{i}.self_attn.out_proj.bias")
+    )
+    rename_keys.append((f"transformer.encoder.layers.{i}.linear1.weight", f"encoder.layers.{i}.fc1.weight"))
+    rename_keys.append((f"transformer.encoder.layers.{i}.linear1.bias", f"encoder.layers.{i}.fc1.bias"))
+    rename_keys.append((f"transformer.encoder.layers.{i}.linear2.weight", f"encoder.layers.{i}.fc2.weight"))
+    rename_keys.append((f"transformer.encoder.layers.{i}.linear2.bias", f"encoder.layers.{i}.fc2.bias"))
+    rename_keys.append(
+        (f"transformer.encoder.layers.{i}.norm1.weight", f"encoder.layers.{i}.self_attn_layer_norm.weight")
+    )
+    rename_keys.append((f"transformer.encoder.layers.{i}.norm1.bias", f"encoder.layers.{i}.self_attn_layer_norm.bias"))
+    rename_keys.append((f"transformer.encoder.layers.{i}.norm2.weight", f"encoder.layers.{i}.final_layer_norm.weight"))
+    rename_keys.append((f"transformer.encoder.layers.{i}.norm2.bias", f"encoder.layers.{i}.final_layer_norm.bias"))
+    # decoder layers: 2 times output projection, 2 feedforward neural networks and 3 layernorms
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.self_attn.out_proj.weight", f"decoder.layers.{i}.self_attn.out_proj.weight")
+    )
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.self_attn.out_proj.bias", f"decoder.layers.{i}.self_attn.out_proj.bias")
+    )
+    rename_keys.append(
+        (
+            f"transformer.decoder.layers.{i}.cross_attn.out_proj.weight",
+            f"decoder.layers.{i}.encoder_attn.out_proj.weight",
+        )
+    )
+    rename_keys.append(
+        (
+            f"transformer.decoder.layers.{i}.cross_attn.out_proj.bias",
+            f"decoder.layers.{i}.encoder_attn.out_proj.bias",
+        )
+    )
+    rename_keys.append((f"transformer.decoder.layers.{i}.linear1.weight", f"decoder.layers.{i}.fc1.weight"))
+    rename_keys.append((f"transformer.decoder.layers.{i}.linear1.bias", f"decoder.layers.{i}.fc1.bias"))
+    rename_keys.append((f"transformer.decoder.layers.{i}.linear2.weight", f"decoder.layers.{i}.fc2.weight"))
+    rename_keys.append((f"transformer.decoder.layers.{i}.linear2.bias", f"decoder.layers.{i}.fc2.bias"))
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.norm1.weight", f"decoder.layers.{i}.self_attn_layer_norm.weight")
+    )
+    rename_keys.append((f"transformer.decoder.layers.{i}.norm1.bias", f"decoder.layers.{i}.self_attn_layer_norm.bias"))
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.norm2.weight", f"decoder.layers.{i}.encoder_attn_layer_norm.weight")
+    )
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.norm2.bias", f"decoder.layers.{i}.encoder_attn_layer_norm.bias")
+    )
+    rename_keys.append((f"transformer.decoder.layers.{i}.norm3.weight", f"decoder.layers.{i}.final_layer_norm.weight"))
+    rename_keys.append((f"transformer.decoder.layers.{i}.norm3.bias", f"decoder.layers.{i}.final_layer_norm.bias"))
+
+    # q, k, v projections in self/cross-attention in decoder for conditional DETR
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.sa_qcontent_proj.weight", f"decoder.layers.{i}.sa_qcontent_proj.weight")
+    )
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.sa_kcontent_proj.weight", f"decoder.layers.{i}.sa_kcontent_proj.weight")
+    )
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.sa_qpos_proj.weight", f"decoder.layers.{i}.sa_qpos_proj.weight")
+    )
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.sa_kpos_proj.weight", f"decoder.layers.{i}.sa_kpos_proj.weight")
+    )
+    rename_keys.append((f"transformer.decoder.layers.{i}.sa_v_proj.weight", f"decoder.layers.{i}.sa_v_proj.weight"))
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.ca_qcontent_proj.weight", f"decoder.layers.{i}.ca_qcontent_proj.weight")
+    )
+    # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.weight", f"decoder.layers.{i}.ca_qpos_proj.weight"))
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.ca_kcontent_proj.weight", f"decoder.layers.{i}.ca_kcontent_proj.weight")
+    )
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.ca_kpos_proj.weight", f"decoder.layers.{i}.ca_kpos_proj.weight")
+    )
+    rename_keys.append((f"transformer.decoder.layers.{i}.ca_v_proj.weight", f"decoder.layers.{i}.ca_v_proj.weight"))
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.ca_qpos_sine_proj.weight", f"decoder.layers.{i}.ca_qpos_sine_proj.weight")
+    )
+
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.sa_qcontent_proj.bias", f"decoder.layers.{i}.sa_qcontent_proj.bias")
+    )
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.sa_kcontent_proj.bias", f"decoder.layers.{i}.sa_kcontent_proj.bias")
+    )
+    rename_keys.append((f"transformer.decoder.layers.{i}.sa_qpos_proj.bias", f"decoder.layers.{i}.sa_qpos_proj.bias"))
+    rename_keys.append((f"transformer.decoder.layers.{i}.sa_kpos_proj.bias", f"decoder.layers.{i}.sa_kpos_proj.bias"))
+    rename_keys.append((f"transformer.decoder.layers.{i}.sa_v_proj.bias", f"decoder.layers.{i}.sa_v_proj.bias"))
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.ca_qcontent_proj.bias", f"decoder.layers.{i}.ca_qcontent_proj.bias")
+    )
+    # rename_keys.append((f"transformer.decoder.layers.{i}.ca_qpos_proj.bias", f"decoder.layers.{i}.ca_qpos_proj.bias"))
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.ca_kcontent_proj.bias", f"decoder.layers.{i}.ca_kcontent_proj.bias")
+    )
+    rename_keys.append((f"transformer.decoder.layers.{i}.ca_kpos_proj.bias", f"decoder.layers.{i}.ca_kpos_proj.bias"))
+    rename_keys.append((f"transformer.decoder.layers.{i}.ca_v_proj.bias", f"decoder.layers.{i}.ca_v_proj.bias"))
+    rename_keys.append(
+        (f"transformer.decoder.layers.{i}.ca_qpos_sine_proj.bias", f"decoder.layers.{i}.ca_qpos_sine_proj.bias")
+    )
+
+# convolutional projection + query embeddings + layernorm of decoder + class and bounding box heads
+# for conditional DETR, also convert reference point head and query scale MLP
+rename_keys.extend(
+    [
+        ("input_proj.weight", "input_projection.weight"),
+        ("input_proj.bias", "input_projection.bias"),
+        ("query_embed.weight", "query_position_embeddings.weight"),
+        ("transformer.decoder.norm.weight", "decoder.layernorm.weight"),
+        ("transformer.decoder.norm.bias", "decoder.layernorm.bias"),
+        ("class_embed.weight", "class_labels_classifier.weight"),
+        ("class_embed.bias", "class_labels_classifier.bias"),
+        ("bbox_embed.layers.0.weight", "bbox_predictor.layers.0.weight"),
+        ("bbox_embed.layers.0.bias", "bbox_predictor.layers.0.bias"),
+        ("bbox_embed.layers.1.weight", "bbox_predictor.layers.1.weight"),
+        ("bbox_embed.layers.1.bias", "bbox_predictor.layers.1.bias"),
+        ("bbox_embed.layers.2.weight", "bbox_predictor.layers.2.weight"),
+        ("bbox_embed.layers.2.bias", "bbox_predictor.layers.2.bias"),
+        ("transformer.decoder.ref_point_head.layers.0.weight", "decoder.ref_point_head.layers.0.weight"),
+        ("transformer.decoder.ref_point_head.layers.0.bias", "decoder.ref_point_head.layers.0.bias"),
+        ("transformer.decoder.ref_point_head.layers.1.weight", "decoder.ref_point_head.layers.1.weight"),
+        ("transformer.decoder.ref_point_head.layers.1.bias", "decoder.ref_point_head.layers.1.bias"),
+        ("transformer.decoder.query_scale.layers.0.weight", "decoder.query_scale.layers.0.weight"),
+        ("transformer.decoder.query_scale.layers.0.bias", "decoder.query_scale.layers.0.bias"),
+        ("transformer.decoder.query_scale.layers.1.weight", "decoder.query_scale.layers.1.weight"),
+        ("transformer.decoder.query_scale.layers.1.bias", "decoder.query_scale.layers.1.bias"),
+        ("transformer.decoder.layers.0.ca_qpos_proj.weight", "decoder.layers.0.ca_qpos_proj.weight"),
+        ("transformer.decoder.layers.0.ca_qpos_proj.bias", "decoder.layers.0.ca_qpos_proj.bias"),
+    ]
+)
+
+
+def rename_key(state_dict, old, new):
+    val = state_dict.pop(old)
+    state_dict[new] = val
+
+
+def rename_backbone_keys(state_dict):
+    new_state_dict = OrderedDict()
+    for key, value in state_dict.items():
+        if "backbone.0.body" in key:
+            new_key = key.replace("backbone.0.body", "backbone.conv_encoder.model")
+            new_state_dict[new_key] = value
+        else:
+            new_state_dict[key] = value
+
+    return new_state_dict
+
+
+def read_in_q_k_v(state_dict, is_panoptic=False):
+    prefix = ""
+    if is_panoptic:
+        prefix = "conditional_detr."
+
+    # first: transformer encoder
+    for i in range(6):
+        # read in weights + bias of input projection layer (in PyTorch's MultiHeadAttention, this is a single matrix + bias)
+        in_proj_weight = state_dict.pop(f"{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_weight")
+        in_proj_bias = state_dict.pop(f"{prefix}transformer.encoder.layers.{i}.self_attn.in_proj_bias")
+        # next, add query, keys and values (in that order) to the state dict
+        state_dict[f"encoder.layers.{i}.self_attn.q_proj.weight"] = in_proj_weight[:256, :]
+        state_dict[f"encoder.layers.{i}.self_attn.q_proj.bias"] = in_proj_bias[:256]
+        state_dict[f"encoder.layers.{i}.self_attn.k_proj.weight"] = in_proj_weight[256:512, :]
+        state_dict[f"encoder.layers.{i}.self_attn.k_proj.bias"] = in_proj_bias[256:512]
+        state_dict[f"encoder.layers.{i}.self_attn.v_proj.weight"] = in_proj_weight[-256:, :]
+        state_dict[f"encoder.layers.{i}.self_attn.v_proj.bias"] = in_proj_bias[-256:]
+
+
+# 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
+
+
+@torch.no_grad()
+def convert_conditional_detr_checkpoint(model_name, pytorch_dump_folder_path):
+    """
+    Copy/paste/tweak model's weights to our CONDITIONAL_DETR structure.
+    """
+
+    # load default config
+    config = ConditionalDetrConfig()
+    # set backbone and dilation attributes
+    if "resnet101" in model_name:
+        config.backbone = "resnet101"
+    if "dc5" in model_name:
+        config.dilation = True
+    is_panoptic = "panoptic" in model_name
+    if is_panoptic:
+        config.num_labels = 250
+    else:
+        config.num_labels = 91
+        repo_id = "datasets/huggingface/label-files"
+        filename = "coco-detection-id2label.json"
+        id2label = json.load(open(hf_hub_download(repo_id, filename), "r"))
+        id2label = {int(k): v for k, v in id2label.items()}
+        config.id2label = id2label
+        config.label2id = {v: k for k, v in id2label.items()}
+
+    # load feature extractor
+    format = "coco_panoptic" if is_panoptic else "coco_detection"
+    feature_extractor = ConditionalDetrFeatureExtractor(format=format)
+
+    # prepare image
+    img = prepare_img()
+    encoding = feature_extractor(images=img, return_tensors="pt")
+    pixel_values = encoding["pixel_values"]
+
+    logger.info(f"Converting model {model_name}...")
+
+    # load original model from torch hub
+    conditional_detr = torch.hub.load("DeppMeng/ConditionalDETR", model_name, pretrained=True).eval()
+    state_dict = conditional_detr.state_dict()
+    # rename keys
+    for src, dest in rename_keys:
+        if is_panoptic:
+            src = "conditional_detr." + src
+        rename_key(state_dict, src, dest)
+    state_dict = rename_backbone_keys(state_dict)
+    # query, key and value matrices need special treatment
+    read_in_q_k_v(state_dict, is_panoptic=is_panoptic)
+    # important: we need to prepend a prefix to each of the base model keys as the head models use different attributes for them
+    prefix = "conditional_detr.model." if is_panoptic else "model."
+    for key in state_dict.copy().keys():
+        if is_panoptic:
+            if (
+                key.startswith("conditional_detr")
+                and not key.startswith("class_labels_classifier")
+                and not key.startswith("bbox_predictor")
+            ):
+                val = state_dict.pop(key)
+                state_dict["conditional_detr.model" + key[4:]] = val
+            elif "class_labels_classifier" in key or "bbox_predictor" in key:
+                val = state_dict.pop(key)
+                state_dict["conditional_detr." + key] = val
+            elif key.startswith("bbox_attention") or key.startswith("mask_head"):
+                continue
+            else:
+                val = state_dict.pop(key)
+                state_dict[prefix + key] = val
+        else:
+            if not key.startswith("class_labels_classifier") and not key.startswith("bbox_predictor"):
+                val = state_dict.pop(key)
+                state_dict[prefix + key] = val
+    # finally, create HuggingFace model and load state dict
+    model = ConditionalDetrForSegmentation(config) if is_panoptic else ConditionalDetrForObjectDetection(config)
+    model.load_state_dict(state_dict)
+    model.eval()
+    model.push_to_hub(repo_id=model_name, organization="DepuMeng", commit_message="Add model")
+    # verify our conversion
+    original_outputs = conditional_detr(pixel_values)
+    outputs = model(pixel_values)
+    assert torch.allclose(outputs.logits, original_outputs["pred_logits"], atol=1e-4)
+    assert torch.allclose(outputs.pred_boxes, original_outputs["pred_boxes"], atol=1e-4)
+    if is_panoptic:
+        assert torch.allclose(outputs.pred_masks, original_outputs["pred_masks"], atol=1e-4)
+
+    # Save model and feature extractor
+    logger.info(f"Saving PyTorch model and feature extractor to {pytorch_dump_folder_path}...")
+    Path(pytorch_dump_folder_path).mkdir(exist_ok=True)
+    model.save_pretrained(pytorch_dump_folder_path)
+    feature_extractor.save_pretrained(pytorch_dump_folder_path)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--model_name",
+        default="conditional_detr_resnet50",
+        type=str,
+        help="Name of the CONDITIONAL_DETR model you'd like to convert.",
+    )
+    parser.add_argument(
+        "--pytorch_dump_folder_path", default=None, type=str, help="Path to the folder to output PyTorch model."
+    )
+    args = parser.parse_args()
+    convert_conditional_detr_checkpoint(args.model_name, args.pytorch_dump_folder_path)

src/transformers/utils/dummy_timm_and_vision_objects.py

@@ -3,6 +3,37 @@
 from ..utils import DummyObject, requires_backends
 
 
+CONDITIONAL_DETR_PRETRAINED_MODEL_ARCHIVE_LIST = None
+
+
+class ConditionalDetrForObjectDetection(metaclass=DummyObject):
+    _backends = ["timm", "vision"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["timm", "vision"])
+
+
+class ConditionalDetrForSegmentation(metaclass=DummyObject):
+    _backends = ["timm", "vision"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["timm", "vision"])
+
+
+class ConditionalDetrModel(metaclass=DummyObject):
+    _backends = ["timm", "vision"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["timm", "vision"])
+
+
+class ConditionalDetrPreTrainedModel(metaclass=DummyObject):
+    _backends = ["timm", "vision"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["timm", "vision"])
+
+
 DEFORMABLE_DETR_PRETRAINED_MODEL_ARCHIVE_LIST = None
 
 

src/transformers/utils/dummy_vision_objects.py

@@ -24,6 +24,13 @@ class CLIPFeatureExtractor(metaclass=DummyObject):
         requires_backends(self, ["vision"])
 
 
+class ConditionalDetrFeatureExtractor(metaclass=DummyObject):
+    _backends = ["vision"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["vision"])
+
+
 class ConvNextFeatureExtractor(metaclass=DummyObject):
     _backends = ["vision"]