[Time-Series] Autoformer model (#21891)

* ran `transformers-cli add-new-model-like` * added `AutoformerLayernorm` and `AutoformerSeriesDecomposition` * added `decomposition_layer` in `init` and `moving_avg` to config * added `AutoformerAutoCorrelation` to encoder & decoder * removed caninical self attention `AutoformerAttention` * added arguments in config and model tester. Init works! 😁 * WIP autoformer attention with autocorrlation * fixed `attn_weights` size * wip time_delay_agg_training * fixing sizes and debug time_delay_agg_training * aggregation in training works! 😁 * `top_k_delays` -> `top_k_delays_index` and added `contiguous()` * wip time_delay_agg_inference * finish time_delay_agg_inference 😎 * added resize to autocorrelation * bug fix: added the length of the output signal to `irfft` * `attention_mask = None` in the decoder * fixed test: changed attention expected size, `test_attention_outputs` works! * removed unnecessary code * apply AutoformerLayernorm in final norm in enc & dec * added series decomposition to the encoder * added series decomp to decoder, with inputs * added trend todos * added autoformer to README * added to index * added autoformer.mdx * remove scaling and init attention_mask in the decoder * make style * fix copies * make fix-copies * inital fix-copies * fix from https://github.com/huggingface/transformers/pull/22076 * make style * fix class names * added trend * added d_model and projection layers * added `trend_projection` source, and decomp layer init * added trend & seasonal init for decoder input * AutoformerModel cannot be copied as it has the decomp layer too * encoder can be copied from time series transformer * fixed generation and made distrb. out more robust * use context window to calculate decomposition * use the context_window for decomposition * use output_params helper * clean up AutoformerAttention * subsequences_length off by 1 * make fix copies * fix test * added init for nn.Conv1d * fix IGNORE_NON_TESTED * added model_doc * fix ruff * ignore tests * remove dup * fix SPECIAL_CASES_TO_ALLOW * do not copy due to conv1d weight init * remove unused imports * added short summary * added label_length and made the model non-autoregressive * added params docs * better doc for `factor` * fix tests * renamed `moving_avg` to `moving_average` * renamed `factor` to `autocorrelation_factor` * make style * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com> * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com> * fix configurations * fix integration tests * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * fixing `lags_sequence` doc * Revert "fixing `lags_sequence` doc" This reverts commit 21e34911e3. * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Apply suggestions from code review Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * model layers now take the config * added `layer_norm_eps` to the config * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * added `config.layer_norm_eps` to AutoformerLayernorm * added `config.layer_norm_eps` to all layernorm layers * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * fix variable names * added inital pretrained model * added use_cache docstring * doc strings for trend and use_cache * fix order of args * imports on one line * fixed get_lagged_subsequences docs * add docstring for create_network_inputs * get rid of layer_norm_eps config * add back layernorm * update fixture location * fix signature * use AutoformerModelOutput dataclass * fix pretrain config * no need as default exists * subclass ModelOutput * remove layer_norm_eps config * fix test_model_outputs_equivalence test * test hidden_states_output * make fix-copies * Update src/transformers/models/autoformer/configuration_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * removed unused attr * Update tests/models/autoformer/test_modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * Update src/transformers/models/autoformer/modeling_autoformer.py Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com> * use AutoFormerDecoderOutput * fix formatting * fix formatting --------- Co-authored-by: Kashif Rasul <kashif.rasul@gmail.com> Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com> Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
2023-05-30 13:53:32 +05:30 · 2023-05-30 13:53:32 +05:30 · 4b6a5a7caa
parent 17a55534f5
commit 4b6a5a7caa
29 changed files with 3058 additions and 16 deletions
--- a/README.md
+++ b/README.md
@ -292,6 +292,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[ALIGN](https://huggingface.co/docs/transformers/model_doc/align)** (from Google Research) released with the paper [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918) by Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig.
 1. **[AltCLIP](https://huggingface.co/docs/transformers/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. **[Autoformer](https://huggingface.co/docs/transformers/main/model_doc/autoformer)** (from Tsinghua University) released with the paper [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
 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.
 1. **[BARTpho](https://huggingface.co/docs/transformers/model_doc/bartpho)** (from VinAI Research) released with the paper [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701) by Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen.
--- a/README_es.md
+++ b/README_es.md
@ -267,6 +267,7 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt
 1. **[ALIGN](https://huggingface.co/docs/transformers/model_doc/align)** (from Google Research) released with the paper [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918) by Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig.
 1. **[AltCLIP](https://huggingface.co/docs/transformers/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. **[Autoformer](https://huggingface.co/docs/transformers/main/model_doc/autoformer)** (from Tsinghua University) released with the paper [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
 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.
 1. **[BARTpho](https://huggingface.co/docs/transformers/model_doc/bartpho)** (from VinAI Research) released with the paper [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701) by Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen.
--- a/README_hd.md
+++ b/README_hd.md
@ -239,6 +239,7 @@ conda install -c huggingface transformers
 1. **[ALIGN](https://huggingface.co/docs/transformers/model_doc/align)** (Google Research से) Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig. द्वाराअनुसंधान पत्र [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918) के साथ जारी किया गया
 1. **[AltCLIP](https://huggingface.co/docs/transformers/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. **[Autoformer](https://huggingface.co/docs/transformers/main/model_doc/autoformer)** (from Tsinghua University) released with the paper [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
 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।
 1. **[BARTpho](https://huggingface.co/docs/transformers/model_doc/bartpho)** (VinAI Research से) साथ में पेपर [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701)गुयेन लुओंग ट्रान, डुओंग मिन्ह ले और डाट क्वोक गुयेन द्वारा पोस्ट किया गया।
--- a/README_ja.md
+++ b/README_ja.md
@ -301,6 +301,7 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ
 1. **[ALIGN](https://huggingface.co/docs/transformers/model_doc/align)** (Google Research から) Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig. から公開された研究論文 [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918)
 1. **[AltCLIP](https://huggingface.co/docs/transformers/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. **[Autoformer](https://huggingface.co/docs/transformers/main/model_doc/autoformer)** (from Tsinghua University) released with the paper [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
 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)
 1. **[BARTpho](https://huggingface.co/docs/transformers/model_doc/bartpho)** (VinAI Research から) Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen から公開された研究論文: [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701)
--- a/README_ko.md
+++ b/README_ko.md
@ -216,6 +216,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[ALIGN](https://huggingface.co/docs/transformers/model_doc/align)** (Google Research 에서 제공)은 Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig.의 [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918)논문과 함께 발표했습니다.
 1. **[AltCLIP](https://huggingface.co/docs/transformers/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. **[Autoformer](https://huggingface.co/docs/transformers/main/model_doc/autoformer)** (from Tsinghua University) released with the paper [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
 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.
 1. **[BARTpho](https://huggingface.co/docs/transformers/model_doc/bartpho)** (from VinAI Research) released with the paper [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701) by Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen.
--- a/README_zh-hans.md
+++ b/README_zh-hans.md
@ -240,6 +240,7 @@ conda install -c huggingface transformers
 1. **[ALIGN](https://huggingface.co/docs/transformers/model_doc/align)** (来自 Google Research) 伴随论文 [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918) 由 Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig 发布。
 1. **[AltCLIP](https://huggingface.co/docs/transformers/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. **[Autoformer](https://huggingface.co/docs/transformers/main/model_doc/autoformer)** (from Tsinghua University) released with the paper [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
 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。
 1. **[BARTpho](https://huggingface.co/docs/transformers/model_doc/bartpho)** (来自 VinAI Research) 伴随论文 [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701) 由 Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen 发布。
--- a/README_zh-hant.md
+++ b/README_zh-hant.md
@ -252,6 +252,7 @@ conda install -c huggingface transformers
 1. **[ALIGN](https://huggingface.co/docs/transformers/model_doc/align)** (from Google Research) released with the paper [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918) by Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig.
 1. **[AltCLIP](https://huggingface.co/docs/transformers/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. **[Autoformer](https://huggingface.co/docs/transformers/main/model_doc/autoformer)** (from Tsinghua University) released with the paper [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
 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.
 1. **[BARTpho](https://huggingface.co/docs/transformers/model_doc/bartpho)** (from VinAI Research) released with the paper [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701) by Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen.
--- a/docs/source/en/_toctree.yml
+++ b/docs/source/en/_toctree.yml
@ -656,6 +656,8 @@
      title: Reinforcement learning models
    - isExpanded: false
      sections:
+      - local: model_doc/autoformer
+        title: Autoformer
      - local: model_doc/informer
        title: Informer
      - local: model_doc/time_series_transformer
--- a/docs/source/en/index.mdx
+++ b/docs/source/en/index.mdx
@ -53,6 +53,7 @@ The documentation is organized into five sections:
 1. **[ALIGN](model_doc/align)** (from Google Research) released with the paper [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918) by Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig.
 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. **[Autoformer](model_doc/autoformer)** (from Tsinghua University) released with the paper [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
 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.
 1. **[BARTpho](model_doc/bartpho)** (from VinAI Research) released with the paper [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701) by Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen.
@ -268,6 +269,7 @@ Flax), PyTorch, and/or TensorFlow.
 |             ALIGN             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            AltCLIP            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 | Audio Spectrogram Transformer |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          Autoformer           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             BART              |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |             BEiT              |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
 |             BERT              |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
--- a/docs/source/en/model_doc/autoformer.mdx
+++ b/docs/source/en/model_doc/autoformer.mdx
@ -0,0 +1,42 @@
+<!--Copyright 2023 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.
+-->
+
+# Autoformer
+
+## Overview
+
+The Autoformer model was proposed in [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://arxiv.org/abs/2106.13008) by Haixu Wu, Jiehui Xu, Jianmin Wang, Mingsheng Long.
+
+This model augments the Transformer as a deep decomposition architecture, which can progressively decompose the trend and seasonal components during the forecasting process.
+
+The abstract from the paper is the following:
+
+*Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-range dependencies. However, intricate temporal patterns of the long-term future prohibit the model from finding reliable dependencies. Also, Transformers have to adopt the sparse versions of point-wise self-attentions for long series efficiency, resulting in the information utilization bottleneck. Going beyond Transformers, we design Autoformer as a novel decomposition architecture with an Auto-Correlation mechanism. We break with the pre-processing convention of series decomposition and renovate it as a basic inner block of deep models. This design empowers Autoformer with progressive decomposition capacities for complex time series. Further, inspired by the stochastic process theory, we design the Auto-Correlation mechanism based on the series periodicity, which conducts the dependencies discovery and representation aggregation at the sub-series level. Auto-Correlation outperforms self-attention in both efficiency and accuracy. In long-term forecasting, Autoformer yields state-of-the-art accuracy, with a 38% relative improvement on six benchmarks, covering five practical applications: energy, traffic, economics, weather and disease.*
+
+This model was contributed by [elisim](https://huggingface.co/elisim) and [kashif](https://huggingface.co/kashif).
+The original code can be found [here](https://github.com/thuml/Autoformer).
+
+## AutoformerConfig
+
+[[autodoc]] AutoformerConfig
+
+
+## AutoformerModel
+
+[[autodoc]] AutoformerModel
+    - forward
+
+
+## AutoformerForPrediction
+
+[[autodoc]] AutoformerForPrediction
+    - forward
--- a/src/transformers/init.py
+++ b/src/transformers/init.py
@ -155,6 +155,10 @@ _import_structure = {
        "AutoProcessor",
        "AutoTokenizer",
    ],
+    "models.autoformer": [
+        "AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP",
+        "AutoformerConfig",
+    ],
    "models.bart": ["BartConfig", "BartTokenizer"],
    "models.barthez": [],
    "models.bartpho": [],
@ -1082,6 +1086,14 @@ else:
            "AutoModelWithLMHead",
        ]
    )
+    _import_structure["models.autoformer"].extend(
+        [
+            "AUTOFORMER_PRETRAINED_MODEL_ARCHIVE_LIST",
+            "AutoformerForPrediction",
+            "AutoformerModel",
+            "AutoformerPreTrainedModel",
+        ]
+    )
    _import_structure["models.bart"].extend(
        [
            "BART_PRETRAINED_MODEL_ARCHIVE_LIST",
@ -3946,6 +3958,10 @@ if TYPE_CHECKING:
        AutoProcessor,
        AutoTokenizer,
    )
+    from .models.autoformer import (
+        AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        AutoformerConfig,
+    )
    from .models.bart import BartConfig, BartTokenizer
    from .models.beit import BEIT_PRETRAINED_CONFIG_ARCHIVE_MAP, BeitConfig
    from .models.bert import (
@ -4784,6 +4800,12 @@ if TYPE_CHECKING:
            AutoModelForZeroShotObjectDetection,
            AutoModelWithLMHead,
        )
+        from .models.autoformer import (
+            AUTOFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
+            AutoformerForPrediction,
+            AutoformerModel,
+            AutoformerPreTrainedModel,
+        )
        from .models.bart import (
            BART_PRETRAINED_MODEL_ARCHIVE_LIST,
            BartForCausalLM,
--- a/src/transformers/models/init.py
+++ b/src/transformers/models/init.py
@ -18,6 +18,7 @@ from . import (
    altclip,
    audio_spectrogram_transformer,
    auto,
+    autoformer,
    bart,
    barthez,
    bartpho,
--- a/src/transformers/models/auto/configuration_auto.py
+++ b/src/transformers/models/auto/configuration_auto.py
@ -33,6 +33,7 @@ CONFIG_MAPPING_NAMES = OrderedDict(
        ("align", "AlignConfig"),
        ("altclip", "AltCLIPConfig"),
        ("audio-spectrogram-transformer", "ASTConfig"),
+        ("autoformer", "AutoformerConfig"),
        ("bart", "BartConfig"),
        ("beit", "BeitConfig"),
        ("bert", "BertConfig"),
@ -225,6 +226,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict(
        ("align", "ALIGN_PRETRAINED_CONFIG_ARCHIVE_MAP"),
        ("altclip", "ALTCLIP_PRETRAINED_CONFIG_ARCHIVE_MAP"),
        ("audio-spectrogram-transformer", "AUDIO_SPECTROGRAM_TRANSFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP"),
+        ("autoformer", "AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP"),
        ("bart", "BART_PRETRAINED_CONFIG_ARCHIVE_MAP"),
        ("beit", "BEIT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
        ("bert", "BERT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
@ -399,6 +401,7 @@ MODEL_NAMES_MAPPING = OrderedDict(
        ("align", "ALIGN"),
        ("altclip", "AltCLIP"),
        ("audio-spectrogram-transformer", "Audio Spectrogram Transformer"),
+        ("autoformer", "Autoformer"),
        ("bart", "BART"),
        ("barthez", "BARThez"),
        ("bartpho", "BARTpho"),
--- a/src/transformers/models/auto/modeling_auto.py
+++ b/src/transformers/models/auto/modeling_auto.py
@ -32,6 +32,7 @@ MODEL_MAPPING_NAMES = OrderedDict(
        ("align", "AlignModel"),
        ("altclip", "AltCLIPModel"),
        ("audio-spectrogram-transformer", "ASTModel"),
+        ("autoformer", "AutoformerModel"),
        ("bart", "BartModel"),
        ("beit", "BeitModel"),
        ("bert", "BertModel"),
--- a/src/transformers/models/autoformer/init.py
+++ b/src/transformers/models/autoformer/init.py
@ -0,0 +1,63 @@
+# Copyright 2023 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_torch_available
+
+
+_import_structure = {
+    "configuration_autoformer": [
+        "AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP",
+        "AutoformerConfig",
+    ],
+}
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_autoformer"] = [
+        "AUTOFORMER_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "AutoformerForPrediction",
+        "AutoformerModel",
+        "AutoformerPreTrainedModel",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_autoformer import (
+        AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        AutoformerConfig,
+    )
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_autoformer import (
+            AUTOFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
+            AutoformerForPrediction,
+            AutoformerModel,
+            AutoformerPreTrainedModel,
+        )
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)
--- a/src/transformers/models/autoformer/configuration_autoformer.py
+++ b/src/transformers/models/autoformer/configuration_autoformer.py
@ -0,0 +1,245 @@
+# coding=utf-8
+# Copyright 2023 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.
+""" Autoformer model configuration"""
+
+from typing import List, Optional
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+AUTOFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "huggingface/autoformer-tourism-monthly": "https://huggingface.co/huggingface/autoformer-tourism-monthly/resolve/main/config.json",
+}
+
+
+class AutoformerConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of an [`AutoformerModel`]. It is used to instantiate an
+    Autoformer 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 Autoformer
+    [huggingface/autoformer-tourism-monthly](https://huggingface.co/huggingface/autoformer-tourism-monthly)
+    architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        prediction_length (`int`):
+            The prediction length for the decoder. In other words, the prediction horizon of the model.
+        context_length (`int`, *optional*, defaults to `prediction_length`):
+            The context length for the encoder. If unset, the context length will be the same as the
+            `prediction_length`.
+        distribution_output (`string`, *optional*, defaults to `"student_t"`):
+            The distribution emission head for the model. Could be either "student_t", "normal" or "negative_binomial".
+        loss (`string`, *optional*, defaults to `"nll"`):
+            The loss function for the model corresponding to the `distribution_output` head. For parametric
+            distributions it is the negative log likelihood (nll) - which currently is the only supported one.
+        input_size (`int`, *optional*, defaults to 1):
+            The size of the target variable which by default is 1 for univariate targets. Would be > 1 in case of
+            multivariate targets.
+        lags_sequence (`list[int]`, *optional*, defaults to `[1, 2, 3, 4, 5, 6, 7]`):
+            The lags of the input time series as covariates often dictated by the frequency. Default is `[1, 2, 3, 4,
+            5, 6, 7]`.
+        scaling (`bool`, *optional* defaults to `True`):
+            Whether to scale the input targets.
+        num_time_features (`int`, *optional*, defaults to 0):
+            The number of time features in the input time series.
+        num_dynamic_real_features (`int`, *optional*, defaults to 0):
+            The number of dynamic real valued features.
+        num_static_categorical_features (`int`, *optional*, defaults to 0):
+            The number of static categorical features.
+        num_static_real_features (`int`, *optional*, defaults to 0):
+            The number of static real valued features.
+        cardinality (`list[int]`, *optional*):
+            The cardinality (number of different values) for each of the static categorical features. Should be a list
+            of integers, having the same length as `num_static_categorical_features`. Cannot be `None` if
+            `num_static_categorical_features` is > 0.
+        embedding_dimension (`list[int]`, *optional*):
+            The dimension of the embedding for each of the static categorical features. Should be a list of integers,
+            having the same length as `num_static_categorical_features`. Cannot be `None` if
+            `num_static_categorical_features` is > 0.
+        d_model (`int`, *optional*, defaults to 64):
+            Dimensionality of the transformer layers.
+        encoder_layers (`int`, *optional*, defaults to 2):
+            Number of encoder layers.
+        decoder_layers (`int`, *optional*, defaults to 2):
+            Number of decoder layers.
+        encoder_attention_heads (`int`, *optional*, defaults to 2):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        decoder_attention_heads (`int`, *optional*, defaults to 2):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        encoder_ffn_dim (`int`, *optional*, defaults to 32):
+            Dimension of the "intermediate" (often named feed-forward) layer in encoder.
+        decoder_ffn_dim (`int`, *optional*, defaults to 32):
+            Dimension of the "intermediate" (often named feed-forward) layer in decoder.
+        activation_function (`str` or `function`, *optional*, defaults to `"gelu"`):
+            The non-linear activation function (function or string) in the encoder and decoder. If string, `"gelu"` and
+            `"relu"` are supported.
+        dropout (`float`, *optional*, defaults to 0.1):
+            The dropout probability for all fully connected layers in the encoder, and decoder.
+        encoder_layerdrop (`float`, *optional*, defaults to 0.1):
+            The dropout probability for the attention and fully connected layers for each encoder layer.
+        decoder_layerdrop (`float`, *optional*, defaults to 0.1):
+            The dropout probability for the attention and fully connected layers for each decoder layer.
+        attention_dropout (`float`, *optional*, defaults to 0.1):
+            The dropout probability for the attention probabilities.
+        activation_dropout (`float`, *optional*, defaults to 0.1):
+            The dropout probability used between the two layers of the feed-forward networks.
+        num_parallel_samples (`int`, *optional*, defaults to 100):
+            The number of samples to generate in parallel for each time step of inference.
+        init_std (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated normal weight initialization distribution.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether to use the past key/values attentions (if applicable to the model) to speed up decoding.
+        label_length (`int`, *optional*, defaults to 10):
+            Start token length of the Autoformer decoder, which is used for direct multi-step prediction (i.e.
+            non-autoregressive generation).
+        moving_average (`int`, defaults to 25):
+            The window size of the moving average. In practice, it's the kernel size in AvgPool1d of the Decomposition
+            Layer.
+        autocorrelation_factor (`int`, defaults to 3):
+            "Attention" (i.e. AutoCorrelation mechanism) factor which is used to find top k autocorrelations delays.
+            It's recommended in the paper to set it to a number between 1 and 5.
+
+
+        Example:
+
+    ```python
+    >>> from transformers import AutoformerConfig, AutoformerModel
+
+    >>> # Initializing a default Autoformer configuration
+    >>> configuration = AutoformerConfig()
+
+    >>> # Randomly initializing a model (with random weights) from the configuration
+    >>> model = AutoformerModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+    model_type = "autoformer"
+    attribute_map = {
+        "hidden_size": "d_model",
+        "num_attention_heads": "encoder_attention_heads",
+        "num_hidden_layers": "encoder_layers",
+    }
+
+    def __init__(
+        self,
+        prediction_length: Optional[int] = None,
+        context_length: Optional[int] = None,
+        distribution_output: str = "student_t",
+        loss: str = "nll",
+        input_size: int = 1,
+        lags_sequence: List[int] = [1, 2, 3, 4, 5, 6, 7],
+        scaling: bool = True,
+        num_time_features: int = 0,
+        num_dynamic_real_features: int = 0,
+        num_static_categorical_features: int = 0,
+        num_static_real_features: int = 0,
+        cardinality: Optional[List[int]] = None,
+        embedding_dimension: Optional[List[int]] = None,
+        d_model: int = 64,
+        encoder_attention_heads: int = 2,
+        decoder_attention_heads: int = 2,
+        encoder_layers: int = 2,
+        decoder_layers: int = 2,
+        encoder_ffn_dim: int = 32,
+        decoder_ffn_dim: int = 32,
+        activation_function: str = "gelu",
+        dropout: float = 0.1,
+        encoder_layerdrop: float = 0.1,
+        decoder_layerdrop: float = 0.1,
+        attention_dropout: float = 0.1,
+        activation_dropout: float = 0.1,
+        num_parallel_samples: int = 100,
+        init_std: float = 0.02,
+        use_cache: bool = True,
+        is_encoder_decoder=True,
+        # Autoformer arguments
+        label_length: int = 10,
+        moving_average: int = 25,
+        autocorrelation_factor: int = 3,
+        **kwargs,
+    ):
+        # time series specific configuration
+        self.prediction_length = prediction_length
+        self.context_length = context_length if context_length is not None else prediction_length
+        self.distribution_output = distribution_output
+        self.loss = loss
+        self.input_size = input_size
+        self.num_time_features = num_time_features
+        self.lags_sequence = lags_sequence
+        self.scaling = scaling
+        self.num_dynamic_real_features = num_dynamic_real_features
+        self.num_static_real_features = num_static_real_features
+        self.num_static_categorical_features = num_static_categorical_features
+        if cardinality is not None and num_static_categorical_features > 0:
+            if len(cardinality) != num_static_categorical_features:
+                raise ValueError(
+                    "The cardinality should be a list of the same length as `num_static_categorical_features`"
+                )
+            self.cardinality = cardinality
+        else:
+            self.cardinality = [0]
+        if embedding_dimension is not None and num_static_categorical_features > 0:
+            if len(embedding_dimension) != num_static_categorical_features:
+                raise ValueError(
+                    "The embedding dimension should be a list of the same length as `num_static_categorical_features`"
+                )
+            self.embedding_dimension = embedding_dimension
+        else:
+            self.embedding_dimension = [min(50, (cat + 1) // 2) for cat in self.cardinality]
+        self.num_parallel_samples = num_parallel_samples
+
+        # Transformer architecture configuration
+        self.feature_size = input_size * len(self.lags_sequence) + self._number_of_features
+        self.d_model = d_model
+        self.encoder_attention_heads = encoder_attention_heads
+        self.decoder_attention_heads = decoder_attention_heads
+        self.encoder_ffn_dim = encoder_ffn_dim
+        self.decoder_ffn_dim = decoder_ffn_dim
+        self.encoder_layers = encoder_layers
+        self.decoder_layers = decoder_layers
+
+        self.dropout = dropout
+        self.attention_dropout = attention_dropout
+        self.activation_dropout = activation_dropout
+        self.encoder_layerdrop = encoder_layerdrop
+        self.decoder_layerdrop = decoder_layerdrop
+
+        self.activation_function = activation_function
+        self.init_std = init_std
+
+        self.use_cache = use_cache
+
+        # Autoformer
+        self.label_length = label_length
+        self.moving_average = moving_average
+        self.autocorrelation_factor = autocorrelation_factor
+
+        super().__init__(is_encoder_decoder=is_encoder_decoder, **kwargs)
+
+    @property
+    def _number_of_features(self) -> int:
+        return (
+            sum(self.embedding_dimension)
+            + self.num_dynamic_real_features
+            + self.num_time_features
+            + self.num_static_real_features
+            + self.input_size * 2  # the log1p(abs(loc)) and log(scale) features
+        )
--- a/src/transformers/models/autoformer/modeling_autoformer.py
+++ b/src/transformers/models/autoformer/modeling_autoformer.py
--- a/src/transformers/models/informer/configuration_informer.py
+++ b/src/transformers/models/informer/configuration_informer.py
@ -232,9 +232,6 @@ class InformerConfig(PretrainedConfig):
        self.activation_function = activation_function
        self.init_std = init_std

-        self.output_attentions = False
-        self.output_hidden_states = False
-
        self.use_cache = use_cache

        # Informer
--- a/src/transformers/models/informer/modeling_informer.py
+++ b/src/transformers/models/informer/modeling_informer.py
@ -142,7 +142,9 @@ class InformerMeanScaler(nn.Module):
        self.default_scale = default_scale

    @torch.no_grad()
-    def forward(self, data: torch.Tensor, observed_indicator: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+    def forward(
+        self, data: torch.Tensor, observed_indicator: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
        # shape: (N, [C], T=1)
        ts_sum = (data * observed_indicator).abs().sum(self.dim, keepdim=True)
        num_observed = observed_indicator.sum(self.dim, keepdim=True)
@ -1669,7 +1671,7 @@ class InformerModel(InformerPreTrainedModel):
        >>> from transformers import InformerModel

        >>> file = hf_hub_download(
-        ...     repo_id="kashif/tourism-monthly-batch", filename="train-batch.pt", repo_type="dataset"
+        ...     repo_id="hf-internal-testing/tourism-monthly-batch", filename="train-batch.pt", repo_type="dataset"
        ... )
        >>> batch = torch.load(file)

@ -1834,7 +1836,7 @@ class InformerForPrediction(InformerPreTrainedModel):
        >>> from transformers import InformerForPrediction

        >>> file = hf_hub_download(
-        ...     repo_id="kashif/tourism-monthly-batch", filename="train-batch.pt", repo_type="dataset"
+        ...     repo_id="hf-internal-testing/tourism-monthly-batch", filename="train-batch.pt", repo_type="dataset"
        ... )
        >>> batch = torch.load(file)

--- a/src/transformers/models/time_series_transformer/configuration_time_series_transformer.py
+++ b/src/transformers/models/time_series_transformer/configuration_time_series_transformer.py
@ -217,9 +217,6 @@ class TimeSeriesTransformerConfig(PretrainedConfig):
        self.activation_function = activation_function
        self.init_std = init_std

-        self.output_attentions = False
-        self.output_hidden_states = False
-
        self.use_cache = use_cache

        super().__init__(is_encoder_decoder=is_encoder_decoder, **kwargs)
--- a/src/transformers/models/time_series_transformer/modeling_time_series_transformer.py
+++ b/src/transformers/models/time_series_transformer/modeling_time_series_transformer.py
@ -140,7 +140,9 @@ class TimeSeriesMeanScaler(nn.Module):
        self.default_scale = default_scale

    @torch.no_grad()
-    def forward(self, data: torch.Tensor, observed_indicator: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+    def forward(
+        self, data: torch.Tensor, observed_indicator: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
        # shape: (N, [C], T=1)
        ts_sum = (data * observed_indicator).abs().sum(self.dim, keepdim=True)
        num_observed = observed_indicator.sum(self.dim, keepdim=True)
@ -1394,7 +1396,7 @@ class TimeSeriesTransformerModel(TimeSeriesTransformerPreTrainedModel):
        >>> from transformers import TimeSeriesTransformerModel

        >>> file = hf_hub_download(
-        ...     repo_id="kashif/tourism-monthly-batch", filename="train-batch.pt", repo_type="dataset"
+        ...     repo_id="hf-internal-testing/tourism-monthly-batch", filename="train-batch.pt", repo_type="dataset"
        ... )
        >>> batch = torch.load(file)

@ -1558,7 +1560,7 @@ class TimeSeriesTransformerForPrediction(TimeSeriesTransformerPreTrainedModel):
        >>> from transformers import TimeSeriesTransformerForPrediction

        >>> file = hf_hub_download(
-        ...     repo_id="kashif/tourism-monthly-batch", filename="train-batch.pt", repo_type="dataset"
+        ...     repo_id="hf-internal-testing/tourism-monthly-batch", filename="train-batch.pt", repo_type="dataset"
        ... )
        >>> batch = torch.load(file)

--- a/src/transformers/time_series_utils.py
+++ b/src/transformers/time_series_utils.py
@ -171,7 +171,7 @@ class StudentTOutput(DistributionOutput):

    @classmethod
    def domain_map(cls, df: torch.Tensor, loc: torch.Tensor, scale: torch.Tensor):
-        scale = cls.squareplus(scale)
+        scale = cls.squareplus(scale).clamp_min(torch.finfo(scale.dtype).eps)
        df = 2.0 + cls.squareplus(df)
        return df.squeeze(-1), loc.squeeze(-1), scale.squeeze(-1)

@ -186,7 +186,7 @@ class NormalOutput(DistributionOutput):

    @classmethod
    def domain_map(cls, loc: torch.Tensor, scale: torch.Tensor):
-        scale = cls.squareplus(scale)
+        scale = cls.squareplus(scale).clamp_min(torch.finfo(scale.dtype).eps)
        return loc.squeeze(-1), scale.squeeze(-1)


--- a/src/transformers/utils/dummy_pt_objects.py
+++ b/src/transformers/utils/dummy_pt_objects.py
@ -772,6 +772,30 @@ class AutoModelWithLMHead(metaclass=DummyObject):
        requires_backends(self, ["torch"])


+AUTOFORMER_PRETRAINED_MODEL_ARCHIVE_LIST = None
+
+
+class AutoformerForPrediction(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class AutoformerModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class AutoformerPreTrainedModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
 BART_PRETRAINED_MODEL_ARCHIVE_LIST = None


--- a/tests/models/autoformer/init.py
+++ b/tests/models/autoformer/init.py
--- a/tests/models/autoformer/test_modeling_autoformer.py
+++ b/tests/models/autoformer/test_modeling_autoformer.py
@ -0,0 +1,449 @@
+# coding=utf-8
+# Copyright 2023 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 Autoformer model. """
+
+import inspect
+import tempfile
+import unittest
+
+from huggingface_hub import hf_hub_download
+
+from transformers import is_torch_available
+from transformers.testing_utils import require_torch, slow, torch_device
+
+from ...test_configuration_common import ConfigTester
+from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
+
+
+TOLERANCE = 1e-4
+
+if is_torch_available():
+    import torch
+
+    from transformers import AutoformerConfig, AutoformerForPrediction, AutoformerModel
+    from transformers.models.autoformer.modeling_autoformer import AutoformerDecoder, AutoformerEncoder
+
+
+@require_torch
+class AutoformerModelTester:
+    def __init__(
+        self,
+        parent,
+        d_model=16,
+        batch_size=13,
+        prediction_length=7,
+        context_length=14,
+        label_length=10,
+        cardinality=19,
+        embedding_dimension=5,
+        num_time_features=4,
+        is_training=True,
+        hidden_size=16,
+        num_hidden_layers=2,
+        num_attention_heads=4,
+        intermediate_size=4,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        lags_sequence=[1, 2, 3, 4, 5],
+        moving_average=25,
+        autocorrelation_factor=5,
+    ):
+        self.d_model = d_model
+        self.parent = parent
+        self.batch_size = batch_size
+        self.prediction_length = prediction_length
+        self.context_length = context_length
+        self.cardinality = cardinality
+        self.num_time_features = num_time_features
+        self.lags_sequence = lags_sequence
+        self.embedding_dimension = embedding_dimension
+        self.is_training = is_training
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.hidden_act = hidden_act
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+
+        self.encoder_seq_length = context_length
+        self.decoder_seq_length = prediction_length + label_length
+        self.label_length = label_length
+
+        self.moving_average = moving_average
+        self.autocorrelation_factor = autocorrelation_factor
+
+    def get_config(self):
+        return AutoformerConfig(
+            d_model=self.d_model,
+            encoder_layers=self.num_hidden_layers,
+            decoder_layers=self.num_hidden_layers,
+            encoder_attention_heads=self.num_attention_heads,
+            decoder_attention_heads=self.num_attention_heads,
+            encoder_ffn_dim=self.intermediate_size,
+            decoder_ffn_dim=self.intermediate_size,
+            dropout=self.hidden_dropout_prob,
+            attention_dropout=self.attention_probs_dropout_prob,
+            prediction_length=self.prediction_length,
+            context_length=self.context_length,
+            label_length=self.label_length,
+            lags_sequence=self.lags_sequence,
+            num_time_features=self.num_time_features,
+            num_static_categorical_features=1,
+            cardinality=[self.cardinality],
+            embedding_dimension=[self.embedding_dimension],
+            moving_average=self.moving_average,
+        )
+
+    def prepare_autoformer_inputs_dict(self, config):
+        _past_length = config.context_length + max(config.lags_sequence)
+
+        static_categorical_features = ids_tensor([self.batch_size, 1], config.cardinality[0])
+        past_time_features = floats_tensor([self.batch_size, _past_length, config.num_time_features])
+        past_values = floats_tensor([self.batch_size, _past_length])
+        past_observed_mask = floats_tensor([self.batch_size, _past_length]) > 0.5
+
+        # decoder inputs
+        future_time_features = floats_tensor([self.batch_size, config.prediction_length, config.num_time_features])
+        future_values = floats_tensor([self.batch_size, config.prediction_length])
+
+        inputs_dict = {
+            "past_values": past_values,
+            "static_categorical_features": static_categorical_features,
+            "past_time_features": past_time_features,
+            "past_observed_mask": past_observed_mask,
+            "future_time_features": future_time_features,
+            "future_values": future_values,
+        }
+        return inputs_dict
+
+    def prepare_config_and_inputs(self):
+        config = self.get_config()
+        inputs_dict = self.prepare_autoformer_inputs_dict(config)
+        return config, inputs_dict
+
+    def prepare_config_and_inputs_for_common(self):
+        config, inputs_dict = self.prepare_config_and_inputs()
+        return config, inputs_dict
+
+    def check_encoder_decoder_model_standalone(self, config, inputs_dict):
+        model = AutoformerModel(config=config).to(torch_device).eval()
+        outputs = model(**inputs_dict)
+
+        encoder_last_hidden_state = outputs.encoder_last_hidden_state
+        last_hidden_state = outputs.last_hidden_state
+
+        with tempfile.TemporaryDirectory() as tmpdirname:
+            encoder = model.get_encoder()
+            encoder.save_pretrained(tmpdirname)
+            encoder = AutoformerEncoder.from_pretrained(tmpdirname).to(torch_device)
+
+        transformer_inputs, feature, _, _, _ = model.create_network_inputs(**inputs_dict)
+        seasonal_input, trend_input = model.decomposition_layer(transformer_inputs[:, : config.context_length, ...])
+
+        enc_input = torch.cat(
+            (transformer_inputs[:, : config.context_length, ...], feature[:, : config.context_length, ...]),
+            dim=-1,
+        )
+        encoder_last_hidden_state_2 = encoder(inputs_embeds=enc_input)[0]
+        self.parent.assertTrue((encoder_last_hidden_state_2 - encoder_last_hidden_state).abs().max().item() < 1e-3)
+
+        mean = (
+            torch.mean(transformer_inputs[:, : config.context_length, ...], dim=1)
+            .unsqueeze(1)
+            .repeat(1, config.prediction_length, 1)
+        )
+        zeros = torch.zeros(
+            [transformer_inputs.shape[0], config.prediction_length, transformer_inputs.shape[2]],
+            device=enc_input.device,
+        )
+
+        dec_input = torch.cat(
+            (
+                torch.cat((seasonal_input[:, -config.label_length :, ...], zeros), dim=1),
+                feature[:, config.context_length - config.label_length :, ...],
+            ),
+            dim=-1,
+        )
+        trend_init = torch.cat(
+            (
+                torch.cat((trend_input[:, -config.label_length :, ...], mean), dim=1),
+                feature[:, config.context_length - config.label_length :, ...],
+            ),
+            dim=-1,
+        )
+
+        with tempfile.TemporaryDirectory() as tmpdirname:
+            decoder = model.get_decoder()
+            decoder.save_pretrained(tmpdirname)
+            decoder = AutoformerDecoder.from_pretrained(tmpdirname).to(torch_device)
+
+        last_hidden_state_2 = decoder(
+            trend=trend_init,
+            inputs_embeds=dec_input,
+            encoder_hidden_states=encoder_last_hidden_state,
+        )[0]
+
+        self.parent.assertTrue((last_hidden_state_2 - last_hidden_state).abs().max().item() < 1e-3)
+
+
+@require_torch
+class AutoformerModelTest(ModelTesterMixin, unittest.TestCase):
+    all_model_classes = (AutoformerModel, AutoformerForPrediction) if is_torch_available() else ()
+    all_generative_model_classes = (AutoformerForPrediction,) if is_torch_available() else ()
+    test_pruning = False
+    test_head_masking = False
+    test_missing_keys = False
+    test_torchscript = False
+    test_inputs_embeds = False
+    test_model_common_attributes = False
+
+    def setUp(self):
+        self.model_tester = AutoformerModelTester(self)
+        self.config_tester = ConfigTester(self, config_class=AutoformerConfig, has_text_modality=False)
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    def test_save_load_strict(self):
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs()
+        for model_class in self.all_model_classes:
+            model = model_class(config)
+
+            with tempfile.TemporaryDirectory() as tmpdirname:
+                model.save_pretrained(tmpdirname)
+                model2, info = model_class.from_pretrained(tmpdirname, output_loading_info=True)
+            self.assertEqual(info["missing_keys"], [])
+
+    def test_encoder_decoder_model_standalone(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs_for_common()
+        self.model_tester.check_encoder_decoder_model_standalone(*config_and_inputs)
+
+    @unittest.skip(reason="Model has no tokens embeddings")
+    def test_resize_tokens_embeddings(self):
+        pass
+
+    # # Input is 'static_categorical_features' not 'input_ids'
+    def test_model_main_input_name(self):
+        model_signature = inspect.signature(getattr(AutoformerModel, "forward"))
+        # The main input is the name of the argument after `self`
+        observed_main_input_name = list(model_signature.parameters.keys())[1]
+        self.assertEqual(AutoformerModel.main_input_name, observed_main_input_name)
+
+    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 = [
+                "past_values",
+                "past_time_features",
+                "past_observed_mask",
+                "static_categorical_features",
+                "static_real_features",
+                "future_values",
+                "future_time_features",
+            ]
+
+            if model.__class__.__name__ in ["AutoformerForPrediction"]:
+                expected_arg_names.append("future_observed_mask")
+
+            expected_arg_names.extend(
+                [
+                    "decoder_attention_mask",
+                    "head_mask",
+                    "decoder_head_mask",
+                    "cross_attn_head_mask",
+                    "encoder_outputs",
+                    "past_key_values",
+                    "output_hidden_states",
+                    "output_attentions",
+                    "use_cache",
+                    "return_dict",
+                ]
+            )
+
+            self.assertListEqual(arg_names[: len(expected_arg_names)], expected_arg_names)
+
+    def test_attention_outputs(self):
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+        config.return_dict = True
+
+        seq_len = getattr(self.model_tester, "seq_length", None)
+        decoder_seq_length = getattr(self.model_tester, "decoder_seq_length", seq_len)
+        encoder_seq_length = getattr(self.model_tester, "encoder_seq_length", seq_len)
+        d_model = getattr(self.model_tester, "d_model", None)
+        num_attention_heads = getattr(self.model_tester, "num_attention_heads", None)
+        dim = d_model // num_attention_heads
+
+        for model_class in self.all_model_classes:
+            inputs_dict["output_attentions"] = True
+            inputs_dict["output_hidden_states"] = False
+            config.return_dict = True
+            model = model_class(config)
+            model.to(torch_device)
+            model.eval()
+            with torch.no_grad():
+                outputs = model(**self._prepare_for_class(inputs_dict, model_class))
+            attentions = outputs.encoder_attentions if config.is_encoder_decoder else outputs.attentions
+            self.assertEqual(len(attentions), self.model_tester.num_hidden_layers)
+
+            # check that output_attentions also work using config
+            del inputs_dict["output_attentions"]
+            config.output_attentions = True
+            model = model_class(config)
+            model.to(torch_device)
+            model.eval()
+            with torch.no_grad():
+                outputs = model(**self._prepare_for_class(inputs_dict, model_class))
+            attentions = outputs.encoder_attentions
+            self.assertEqual(len(attentions), self.model_tester.num_hidden_layers)
+
+            self.assertListEqual(
+                list(attentions[0].shape[-3:]),
+                [self.model_tester.num_attention_heads, encoder_seq_length, dim],
+            )
+            out_len = len(outputs)
+
+            correct_outlen = 7
+
+            if "last_hidden_state" in outputs:
+                correct_outlen += 1
+
+            if "trend" in outputs:
+                correct_outlen += 1
+
+            if "past_key_values" in outputs:
+                correct_outlen += 1  # past_key_values have been returned
+
+            if "loss" in outputs:
+                correct_outlen += 1
+
+            if "params" in outputs:
+                correct_outlen += 1
+
+            self.assertEqual(out_len, correct_outlen)
+
+            # decoder attentions
+            decoder_attentions = outputs.decoder_attentions
+            self.assertIsInstance(decoder_attentions, (list, tuple))
+            self.assertEqual(len(decoder_attentions), self.model_tester.num_hidden_layers)
+            self.assertListEqual(
+                list(decoder_attentions[0].shape[-3:]),
+                [self.model_tester.num_attention_heads, decoder_seq_length, dim],
+            )
+
+            # cross attentions
+            cross_attentions = outputs.cross_attentions
+            self.assertIsInstance(cross_attentions, (list, tuple))
+            self.assertEqual(len(cross_attentions), self.model_tester.num_hidden_layers)
+            self.assertListEqual(
+                list(cross_attentions[0].shape[-3:]),
+                [self.model_tester.num_attention_heads, decoder_seq_length, dim],
+            )
+
+        # Check attention is always last and order is fine
+        inputs_dict["output_attentions"] = True
+        inputs_dict["output_hidden_states"] = True
+        model = model_class(config)
+        model.to(torch_device)
+        model.eval()
+        with torch.no_grad():
+            outputs = model(**self._prepare_for_class(inputs_dict, model_class))
+
+        self.assertEqual(out_len + 2, len(outputs))
+
+        self_attentions = outputs.encoder_attentions if config.is_encoder_decoder else outputs.attentions
+
+        self.assertEqual(len(self_attentions), self.model_tester.num_hidden_layers)
+        self.assertListEqual(
+            list(self_attentions[0].shape[-3:]),
+            [self.model_tester.num_attention_heads, encoder_seq_length, dim],
+        )
+
+
+def prepare_batch(filename="train-batch.pt"):
+    file = hf_hub_download(repo_id="hf-internal-testing/tourism-monthly-batch", filename=filename, repo_type="dataset")
+    batch = torch.load(file, map_location=torch_device)
+    return batch
+
+
+@require_torch
+@slow
+class AutoformerModelIntegrationTests(unittest.TestCase):
+    def test_inference_no_head(self):
+        model = AutoformerModel.from_pretrained("huggingface/autoformer-tourism-monthly").to(torch_device)
+        batch = prepare_batch()
+
+        with torch.no_grad():
+            output = model(
+                past_values=batch["past_values"],
+                past_time_features=batch["past_time_features"],
+                past_observed_mask=batch["past_observed_mask"],
+                static_categorical_features=batch["static_categorical_features"],
+                future_values=batch["future_values"],
+                future_time_features=batch["future_time_features"],
+            )[0]
+
+        expected_shape = torch.Size(
+            (64, model.config.prediction_length + model.config.label_length, model.config.feature_size)
+        )
+        self.assertEqual(output.shape, expected_shape)
+
+        expected_slice = torch.tensor(
+            [[0.3593, -1.3398, 0.6330], [0.2279, 1.5396, -0.1792], [0.0450, 1.3225, -0.2335]], device=torch_device
+        )
+        self.assertTrue(torch.allclose(output[0, :3, :3], expected_slice, atol=TOLERANCE))
+
+    def test_inference_head(self):
+        model = AutoformerForPrediction.from_pretrained("huggingface/autoformer-tourism-monthly").to(torch_device)
+        batch = prepare_batch("val-batch.pt")
+        with torch.no_grad():
+            output = model(
+                past_values=batch["past_values"],
+                past_time_features=batch["past_time_features"],
+                past_observed_mask=batch["past_observed_mask"],
+                static_categorical_features=batch["static_categorical_features"],
+            ).encoder_last_hidden_state
+        expected_shape = torch.Size((64, model.config.context_length, model.config.d_model))
+        self.assertEqual(output.shape, expected_shape)
+
+        expected_slice = torch.tensor(
+            [[-0.0734, -0.9036, 0.8358], [4.7186, 2.4113, 1.9581], [1.7953, 2.3558, 1.2970]], device=torch_device
+        )
+        self.assertTrue(torch.allclose(output[0, :3, :3], expected_slice, atol=TOLERANCE))
+
+    def test_seq_to_seq_generation(self):
+        model = AutoformerForPrediction.from_pretrained("huggingface/autoformer-tourism-monthly").to(torch_device)
+        batch = prepare_batch("val-batch.pt")
+        with torch.no_grad():
+            outputs = model.generate(
+                static_categorical_features=batch["static_categorical_features"],
+                past_time_features=batch["past_time_features"],
+                past_values=batch["past_values"],
+                future_time_features=batch["future_time_features"],
+                past_observed_mask=batch["past_observed_mask"],
+            )
+        expected_shape = torch.Size((64, model.config.num_parallel_samples, model.config.prediction_length))
+        self.assertEqual(outputs.sequences.shape, expected_shape)
+
+        expected_slice = torch.tensor([3130.6763, 4056.5293, 7053.0786], device=torch_device)
+        mean_prediction = outputs.sequences.mean(dim=1)
+        self.assertTrue(torch.allclose(mean_prediction[0, -3:], expected_slice, rtol=1e-1))
--- a/tests/models/informer/test_modeling_informer.py
+++ b/tests/models/informer/test_modeling_informer.py
@ -438,7 +438,7 @@ class InformerModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase


 def prepare_batch(filename="train-batch.pt"):
-    file = hf_hub_download(repo_id="kashif/tourism-monthly-batch", filename=filename, repo_type="dataset")
+    file = hf_hub_download(repo_id="hf-internal-testing/tourism-monthly-batch", filename=filename, repo_type="dataset")
    batch = torch.load(file, map_location=torch_device)
    return batch

--- a/tests/models/time_series_transformer/test_modeling_time_series_transformer.py
+++ b/tests/models/time_series_transformer/test_modeling_time_series_transformer.py
@ -459,7 +459,7 @@ class TimeSeriesTransformerModelTest(ModelTesterMixin, PipelineTesterMixin, unit


 def prepare_batch(filename="train-batch.pt"):
-    file = hf_hub_download(repo_id="kashif/tourism-monthly-batch", filename=filename, repo_type="dataset")
+    file = hf_hub_download(repo_id="hf-internal-testing/tourism-monthly-batch", filename=filename, repo_type="dataset")
    batch = torch.load(file, map_location=torch_device)
    return batch

--- a/utils/check_config_attributes.py
+++ b/utils/check_config_attributes.py
@ -73,6 +73,8 @@ SPECIAL_CASES_TO_ALLOW = {
    "InformerConfig": ["num_static_real_features", "num_time_features"],
    # used internally to calculate the feature size
    "TimeSeriesTransformerConfig": ["num_static_real_features", "num_time_features"],
+    # used internally to calculate the feature size
+    "AutoformerConfig": ["num_static_real_features", "num_time_features"],
 }

 # TODO (ydshieh): Check the failing cases, try to fix them or move some cases to the above block once we are sure
--- a/utils/check_repo.py
+++ b/utils/check_repo.py
@ -73,6 +73,8 @@ IGNORE_NON_TESTED = PRIVATE_MODELS.copy() + [
    "TimeSeriesTransformerDecoder",  # Building part of bigger (tested) model.
    "InformerEncoder",  # Building part of bigger (tested) model.
    "InformerDecoder",  # Building part of bigger (tested) model.
+    "AutoformerEncoder",  # Building part of bigger (tested) model.
+    "AutoformerDecoder",  # Building part of bigger (tested) model.
    "JukeboxVQVAE",  # Building part of bigger (tested) model.
    "JukeboxPrior",  # Building part of bigger (tested) model.
    "DeformableDetrEncoder",  # Building part of bigger (tested) model.
@ -223,6 +225,7 @@ IGNORE_NON_AUTO_CONFIGURED = PRIVATE_MODELS.copy() + [
    "GPTSanJapaneseModel",
    "TimeSeriesTransformerForPrediction",
    "InformerForPrediction",
+    "AutoformerForPrediction",
    "JukeboxVQVAE",
    "JukeboxPrior",
    "PegasusXEncoder",