[WIP] add SpeechT5 model (#18922)

* make SpeechT5 model by copying Wav2Vec2

* add paper to docs

* whoops added docs in wrong file

* remove SpeechT5Tokenizer + put CTC back in the name

* remove deprecated class

* remove unused docstring

* delete SpeechT5FeatureExtractor, use Wav2Vec2FeatureExtractor instead

* remove classes we don't need right now

* initial stab at speech encoder prenet

* add more speech encoder prenet stuff

* improve SpeechEncoderPrenet

* add encoder (not finished yet)

* add relative position bias to self-attention

* add encoder CTC layers

* fix formatting

* add decoder from BART, doesn't work yet

* make it work with generate loop

* wrap the encoder into a speech encoder class

* wrap the decoder in a text decoder class

* changed my mind

* changed my mind again ;-)

* load decoder weights, make it work

* add weights for text decoder postnet

* add SpeechT5ForCTC model that uses only the encoder

* clean up EncoderLayer and DecoderLayer

* implement _init_weights in SpeechT5PreTrainedModel

* cleanup config + Encoder and Decoder

* add head + cross attention masks

* improve doc comments

* fixup

* more cleanup

* more fixup

* TextDecoderPrenet works now, thanks Kendall

* add CTC loss

* add placeholders for other pre/postnets

* add type annotation

* fix freeze_feature_encoder

* set padding tokens to 0 in decoder attention mask

* encoder attention mask downsampling

* remove features_pen calculation

* disable the padding tokens thing again

* fixup

* more fixup

* code review fixes

* rename encoder/decoder wrapper classes

* allow checkpoints to be loaded into SpeechT5Model

* put encoder into wrapper for CTC model

* clean up conversion script

* add encoder for TTS model

* add speech decoder prenet

* add speech decoder post-net

* attempt to reconstruct the generation loop

* add speech generation loop

* clean up generate_speech

* small tweaks

* fix forward pass

* enable always dropout on speech decoder prenet

* sort declaration

* rename models

* fixup

* fix copies

* more fixup

* make consistency checker happy

* add Seq2SeqSpectrogramOutput class

* doc comments

* quick note about loss and labels

* add HiFi-GAN implementation (from Speech2Speech PR)

* rename file

* add vocoder to TTS model

* improve vocoder

* working on tokenizer

* more better tokenizer

* add CTC tokenizer

* fix decode and batch_code in CTC tokenizer

* fix processor

* two processors and feature extractors

* use SpeechT5WaveformFeatureExtractor instead of Wav2Vec2

* cleanup

* more cleanup

* even more fixup

* notebooks

* fix log-mel spectrograms

* support reduction factor

* fixup

* shift spectrograms to right to create decoder inputs

* return correct labels

* add labels for stop token prediction

* fix doc comments

* fixup

* remove SpeechT5ForPreTraining

* more fixup

* update copyright headers

* add usage examples

* add SpeechT5ProcessorForCTC

* fixup

* push unofficial checkpoints to hub

* initial version of tokenizer unit tests

* add slow test

* fix failing tests

* tests for CTC tokenizer

* finish CTC tokenizer tests

* processor tests

* initial test for feature extractors

* tests for spectrogram feature extractor

* fixup

* more fixup

* add decorators

* require speech for tests

* modeling tests

* more tests for ASR model

* fix imports

* add fake tests for the other models

* fixup

* remove jupyter notebooks

* add missing SpeechT5Model tests

* add missing tests for SpeechT5ForCTC

* add missing tests for SpeechT5ForTextToSpeech

* sort tests by name

* fix Hi-Fi GAN tests

* fixup

* add speech-to-speech model

* refactor duplicate speech generation code

* add processor for SpeechToSpeech model

* add usage example

* add tests for speech-to-speech model

* fixup

* enable gradient checkpointing for SpeechT5FeatureEncoder

* code review

* push_to_hub now takes repo_id

* improve doc comments for HiFi-GAN config

* add missing test

* add integration tests

* make number of layers in speech decoder prenet configurable

* rename variable

* rename variables

* add auto classes for TTS and S2S

* REMOVE CTC!!!

* S2S processor does not support save/load_pretrained

* fixup

* these models are now in an auto mapping

* fix doc links

* rename HiFiGAN to HifiGan, remove separate config file

* REMOVE auto classes

* there can be only one

* fixup

* replace assert

* reformat

* feature extractor can process input and target at same time

* update checkpoint names

* fix commit hash

README.md

@@ -400,6 +400,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (from NVIDIA) released with the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo.
 1. **[SEW](https://huggingface.co/docs/transformers/model_doc/sew)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
 1. **[SEW-D](https://huggingface.co/docs/transformers/model_doc/sew_d)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
+1. **[SpeechT5](https://huggingface.co/docs/transformers/main/model_doc/speecht5)** (from Microsoft Research) released with the paper [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205) by Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei.
 1. **[SpeechToTextTransformer](https://huggingface.co/docs/transformers/model_doc/speech_to_text)** (from Facebook), released together with the paper [fairseq S2T: Fast Speech-to-Text Modeling with fairseq](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino.
 1. **[SpeechToTextTransformer2](https://huggingface.co/docs/transformers/model_doc/speech_to_text_2)** (from Facebook), released together with the paper [Large-Scale Self- and Semi-Supervised Learning for Speech Translation](https://arxiv.org/abs/2104.06678) by Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau.
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (from Tel Aviv University), released together with the paper [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy.

README_es.md

@@ -393,6 +393,7 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (from NVIDIA) released with the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo.
 1. **[SEW](https://huggingface.co/docs/transformers/model_doc/sew)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
 1. **[SEW-D](https://huggingface.co/docs/transformers/model_doc/sew_d)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
+1. **[SpeechT5](https://huggingface.co/docs/transformers/main/model_doc/speecht5)** (from Microsoft Research) released with the paper [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205) by Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei.
 1. **[SpeechToTextTransformer](https://huggingface.co/docs/transformers/model_doc/speech_to_text)** (from Facebook), released together with the paper [fairseq S2T: Fast Speech-to-Text Modeling with fairseq](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino.
 1. **[SpeechToTextTransformer2](https://huggingface.co/docs/transformers/model_doc/speech_to_text_2)** (from Facebook), released together with the paper [Large-Scale Self- and Semi-Supervised Learning for Speech Translation](https://arxiv.org/abs/2104.06678) by Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau.
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (from Tel Aviv University), released together with the paper [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy.

README_hd.md

@@ -365,6 +365,7 @@ conda install -c huggingface transformers
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (from NVIDIA) released with the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo.
 1. **[SEW](https://huggingface.co/docs/transformers/model_doc/sew)** (ASAPP से) साथ देने वाला पेपर [भाषण पहचान के लिए अनसुपरवाइज्ड प्री-ट्रेनिंग में परफॉर्मेंस-एफिशिएंसी ट्रेड-ऑफ्स](https ://arxiv.org/abs/2109.06870) फेलिक्स वू, क्वांगयुन किम, जिंग पैन, क्यू हान, किलियन क्यू. वेनबर्गर, योव आर्टज़ी द्वारा।
 1. **[SEW-D](https://huggingface.co/docs/transformers/model_doc/sew_d)** (ASAPP से) साथ में पेपर [भाषण पहचान के लिए अनसुपरवाइज्ड प्री-ट्रेनिंग में परफॉर्मेंस-एफिशिएंसी ट्रेड-ऑफ्स] (https://arxiv.org/abs/2109.06870) फेलिक्स वू, क्वांगयुन किम, जिंग पैन, क्यू हान, किलियन क्यू. वेनबर्गर, योआव आर्टज़ी द्वारा पोस्ट किया गया।
+1. **[SpeechT5](https://huggingface.co/docs/transformers/main/model_doc/speecht5)** (from Microsoft Research) released with the paper [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205) by Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei.
 1. **[SpeechToTextTransformer](https://huggingface.co/docs/transformers/model_doc/speech_to_text)** (फेसबुक से), साथ में पेपर [फेयरसेक S2T: फास्ट स्पीच-टू-टेक्स्ट मॉडलिंग विद फेयरसेक](https: //arxiv.org/abs/2010.05171) चांगहान वांग, यूं तांग, जुताई मा, ऐनी वू, दिमित्रो ओखोनको, जुआन पिनो द्वारा पोस्ट किया गया。
 1. **[SpeechToTextTransformer2](https://huggingface.co/docs/transformers/model_doc/speech_to_text_2)** (फेसबुक से) साथ में पेपर [लार्ज-स्केल सेल्फ- एंड सेमी-सुपरवाइज्ड लर्निंग फॉर स्पीच ट्रांसलेशन](https://arxiv.org/abs/2104.06678) चांगहान वांग, ऐनी वू, जुआन पिनो, एलेक्सी बेवस्की, माइकल औली, एलेक्सिस द्वारा Conneau द्वारा पोस्ट किया गया।
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (तेल अवीव यूनिवर्सिटी से) साथ में पेपर [स्पैन सिलेक्शन को प्री-ट्रेनिंग करके कुछ-शॉट क्वेश्चन आंसरिंग](https:// arxiv.org/abs/2101.00438) ओरि राम, युवल कर्स्टन, जोनाथन बेरेंट, अमीर ग्लोबर्सन, ओमर लेवी द्वारा।

README_ja.md

@@ -427,6 +427,7 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (NVIDIA から) Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo から公開された研究論文: [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203)
 1. **[SEW](https://huggingface.co/docs/transformers/model_doc/sew)** (ASAPP から) Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi から公開された研究論文: [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870)
 1. **[SEW-D](https://huggingface.co/docs/transformers/model_doc/sew_d)** (ASAPP から) Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi から公開された研究論文: [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870)
+1. **[SpeechT5](https://huggingface.co/docs/transformers/main/model_doc/speecht5)** (Microsoft Research から) Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei. から公開された研究論文 [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205)
 1. **[SpeechToTextTransformer](https://huggingface.co/docs/transformers/model_doc/speech_to_text)** (Facebook から), Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino から公開された研究論文: [fairseq S2T: Fast Speech-to-Text Modeling with fairseq](https://arxiv.org/abs/2010.05171)
 1. **[SpeechToTextTransformer2](https://huggingface.co/docs/transformers/model_doc/speech_to_text_2)** (Facebook から), Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau から公開された研究論文: [Large-Scale Self- and Semi-Supervised Learning for Speech Translation](https://arxiv.org/abs/2104.06678)
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (Tel Aviv University から), Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy から公開された研究論文: [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438)

README_ko.md

@@ -342,6 +342,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (NVIDIA 에서) Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo 의 [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) 논문과 함께 발표했습니다.
 1. **[SEW](https://huggingface.co/docs/transformers/model_doc/sew)** (ASAPP 에서) Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi 의 [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) 논문과 함께 발표했습니다.
 1. **[SEW-D](https://huggingface.co/docs/transformers/model_doc/sew_d)** (ASAPP 에서) Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi 의 [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) 논문과 함께 발표했습니다.
+1. **[SpeechT5](https://huggingface.co/docs/transformers/main/model_doc/speecht5)** (Microsoft Research 에서 제공)은 Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei.의 [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205)논문과 함께 발표했습니다.
 1. **[SpeechToTextTransformer](https://huggingface.co/docs/transformers/model_doc/speech_to_text)** (Facebook 에서) Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino 의 [fairseq S2T: Fast Speech-to-Text Modeling with fairseq](https://arxiv.org/abs/2010.05171) 논문과 함께 발표했습니다.
 1. **[SpeechToTextTransformer2](https://huggingface.co/docs/transformers/model_doc/speech_to_text_2)** (Facebook 에서) Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau 의 [Large-Scale Self- and Semi-Supervised Learning for Speech Translation](https://arxiv.org/abs/2104.06678) 논문과 함께 발표했습니다.
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (Tel Aviv University 에서) Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy 의 [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) 논문과 함께 발표했습니다.

README_zh-hans.md

@@ -366,6 +366,7 @@ conda install -c huggingface transformers
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (来自 NVIDIA) 伴随论文 [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) 由 Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo 发布。
 1. **[SEW](https://huggingface.co/docs/transformers/model_doc/sew)** (来自 ASAPP) 伴随论文 [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) 由 Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi 发布。
 1. **[SEW-D](https://huggingface.co/docs/transformers/model_doc/sew_d)** (来自 ASAPP) 伴随论文 [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) 由 Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi 发布。
+1. **[SpeechT5](https://huggingface.co/docs/transformers/main/model_doc/speecht5)** (来自 Microsoft Research) 伴随论文 [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205) 由 Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei 发布。
 1. **[SpeechToTextTransformer](https://huggingface.co/docs/transformers/model_doc/speech_to_text)** (来自 Facebook), 伴随论文 [fairseq S2T: Fast Speech-to-Text Modeling with fairseq](https://arxiv.org/abs/2010.05171) 由 Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino 发布。
 1. **[SpeechToTextTransformer2](https://huggingface.co/docs/transformers/model_doc/speech_to_text_2)** (来自 Facebook) 伴随论文 [Large-Scale Self- and Semi-Supervised Learning for Speech Translation](https://arxiv.org/abs/2104.06678) 由 Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau 发布。
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (来自 Tel Aviv University) 伴随论文 [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) 由 Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy 发布。

README_zh-hant.md

@@ -378,6 +378,7 @@ conda install -c huggingface transformers
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (from NVIDIA) released with the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo.
 1. **[SEW](https://huggingface.co/docs/transformers/model_doc/sew)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
 1. **[SEW-D](https://huggingface.co/docs/transformers/model_doc/sew_d)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
+1. **[SpeechT5](https://huggingface.co/docs/transformers/main/model_doc/speecht5)** (from Microsoft Research) released with the paper [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205) by Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei.
 1. **[SpeechToTextTransformer](https://huggingface.co/docs/transformers/model_doc/speech_to_text)** (from Facebook), released together with the paper [fairseq S2T: Fast Speech-to-Text Modeling with fairseq](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino.
 1. **[SpeechToTextTransformer2](https://huggingface.co/docs/transformers/model_doc/speech_to_text_2)** (from Facebook) released with the paper [Large-Scale Self- and Semi-Supervised Learning for Speech Translation](https://arxiv.org/abs/2104.06678) by Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau.
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (from Tel Aviv University) released with the paper [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy.

docs/source/en/_toctree.yml

@@ -491,6 +491,8 @@
         title: Speech2Text
       - local: model_doc/speech_to_text_2
         title: Speech2Text2
+      - local: model_doc/speecht5
+        title: SpeechT5
       - local: model_doc/unispeech
         title: UniSpeech
       - local: model_doc/unispeech-sat

docs/source/en/index.mdx

@@ -179,6 +179,7 @@ The documentation is organized into five sections:
 1. **[SegFormer](model_doc/segformer)** (from NVIDIA) released with the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo.
 1. **[SEW](model_doc/sew)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
 1. **[SEW-D](model_doc/sew_d)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
+1. **[SpeechT5](model_doc/speecht5)** (from Microsoft Research) released with the paper [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205) by Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei.
 1. **[SpeechToTextTransformer](model_doc/speech_to_text)** (from Facebook), released together with the paper [fairseq S2T: Fast Speech-to-Text Modeling with fairseq](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino.
 1. **[SpeechToTextTransformer2](model_doc/speech_to_text_2)** (from Facebook), released together with the paper [Large-Scale Self- and Semi-Supervised Learning for Speech Translation](https://arxiv.org/abs/2104.06678) by Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau.
 1. **[Splinter](model_doc/splinter)** (from Tel Aviv University), released together with the paper [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy.
@@ -360,6 +361,7 @@ Flax), PyTorch, and/or TensorFlow.
 |    Speech Encoder decoder     |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
 |          Speech2Text          |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
 |         Speech2Text2          |       ✅       |       ❌       |       ❌        |         ❌         |      ❌      |
+|           SpeechT5            |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |           Splinter            |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
 |          SqueezeBERT          |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
 |       Swin Transformer        |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |

docs/source/en/main_classes/output.mdx

@@ -136,6 +136,10 @@ documented on their corresponding model page.
 
 [[autodoc]] modeling_outputs.Seq2SeqQuestionAnsweringModelOutput
 
+## Seq2SeqSpectrogramOutput
+
+[[autodoc]] modeling_outputs.Seq2SeqSpectrogramOutput
+
 ## SemanticSegmenterOutput
 
 [[autodoc]] modeling_outputs.SemanticSegmenterOutput

docs/source/en/model_doc/speecht5.mdx

+<!--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.
+-->
+
+# SpeechT5
+
+## Overview
+
+The SpeechT5 model was proposed in [SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing](https://arxiv.org/abs/2110.07205) by Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei.
+
+The abstract from the paper is the following:
+
+*Motivated by the success of T5 (Text-To-Text Transfer Transformer) in pre-trained natural language processing models, we propose a unified-modal SpeechT5 framework that explores the encoder-decoder pre-training for self-supervised speech/text representation learning. The SpeechT5 framework consists of a shared encoder-decoder network and six modal-specific (speech/text) pre/post-nets. After preprocessing the input speech/text through the pre-nets, the shared encoder-decoder network models the sequence-to-sequence transformation, and then the post-nets generate the output in the speech/text modality based on the output of the decoder. Leveraging large-scale unlabeled speech and text data, we pre-train SpeechT5 to learn a unified-modal representation, hoping to improve the modeling capability for both speech and text. To align the textual and speech information into this unified semantic space, we propose a cross-modal vector quantization approach that randomly mixes up speech/text states with latent units as the interface between encoder and decoder. Extensive evaluations show the superiority of the proposed SpeechT5 framework on a wide variety of spoken language processing tasks, including automatic speech recognition, speech synthesis, speech translation, voice conversion, speech enhancement, and speaker identification.*
+
+This model was contributed by [Matthijs](https://huggingface.co/Matthijs). The original code can be found [here](https://github.com/microsoft/SpeechT5).
+
+## SpeechT5Config
+
+[[autodoc]] SpeechT5Config
+
+## SpeechT5HifiGanConfig
+
+[[autodoc]] SpeechT5HifiGanConfig
+
+## SpeechT5Tokenizer
+
+[[autodoc]] SpeechT5Tokenizer
+    - __call__
+    - save_vocabulary
+    - decode
+    - batch_decode
+
+## SpeechT5FeatureExtractor
+
+[[autodoc]] SpeechT5FeatureExtractor
+    - __call__
+
+## SpeechT5Processor
+
+[[autodoc]] SpeechT5Processor
+    - __call__
+    - pad
+    - from_pretrained
+    - save_pretrained
+    - batch_decode
+    - decode
+
+## SpeechT5Model
+
+[[autodoc]] SpeechT5Model
+    - forward
+
+## SpeechT5ForSpeechToText
+
+[[autodoc]] SpeechT5ForSpeechToText
+    - forward
+
+## SpeechT5ForTextToSpeech
+
+[[autodoc]] SpeechT5ForTextToSpeech
+    - forward
+    - generate_speech
+
+## SpeechT5ForSpeechToSpeech
+
+[[autodoc]] SpeechT5ForSpeechToSpeech
+    - forward
+    - generate_speech
+
+## SpeechT5HifiGan
+
+[[autodoc]] SpeechT5HifiGan
+    - forward

src/transformers/__init__.py

@@ -402,6 +402,12 @@ _import_structure = {
         "Speech2Text2Processor",
         "Speech2Text2Tokenizer",
     ],
+    "models.speecht5": [
+        "SPEECHT5_PRETRAINED_CONFIG_ARCHIVE_MAP",
+        "SPEECHT5_PRETRAINED_HIFIGAN_CONFIG_ARCHIVE_MAP",
+        "SpeechT5Config",
+        "SpeechT5HifiGanConfig",
+    ],
     "models.splinter": ["SPLINTER_PRETRAINED_CONFIG_ARCHIVE_MAP", "SplinterConfig", "SplinterTokenizer"],
     "models.squeezebert": ["SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "SqueezeBertConfig", "SqueezeBertTokenizer"],
     "models.swin": ["SWIN_PRETRAINED_CONFIG_ARCHIVE_MAP", "SwinConfig"],
@@ -632,6 +638,7 @@ else:
     _import_structure["models.reformer"].append("ReformerTokenizer")
     _import_structure["models.rembert"].append("RemBertTokenizer")
     _import_structure["models.speech_to_text"].append("Speech2TextTokenizer")
+    _import_structure["models.speecht5"].append("SpeechT5Tokenizer")
     _import_structure["models.t5"].append("T5Tokenizer")
     _import_structure["models.xglm"].append("XGLMTokenizer")
     _import_structure["models.xlm_prophetnet"].append("XLMProphetNetTokenizer")
@@ -734,6 +741,7 @@ else:
     _import_structure["models.audio_spectrogram_transformer"].append("ASTFeatureExtractor")
     _import_structure["models.mctct"].append("MCTCTFeatureExtractor")
     _import_structure["models.speech_to_text"].append("Speech2TextFeatureExtractor")
+    _import_structure["models.speecht5"].append("SpeechT5FeatureExtractor")
 
 # Tensorflow-text-specific objects
 try:
@@ -772,6 +780,7 @@ except OptionalDependencyNotAvailable:
     ]
 else:
     _import_structure["models.speech_to_text"].append("Speech2TextProcessor")
+    _import_structure["models.speecht5"].append("SpeechT5Processor")
 
 # Vision-specific objects
 try:
@@ -2193,6 +2202,17 @@ else:
         ]
     )
     _import_structure["models.speech_to_text_2"].extend(["Speech2Text2ForCausalLM", "Speech2Text2PreTrainedModel"])
+    _import_structure["models.speecht5"].extend(
+        [
+            "SPEECHT5_PRETRAINED_MODEL_ARCHIVE_LIST",
+            "SpeechT5ForSpeechToSpeech",
+            "SpeechT5ForSpeechToText",
+            "SpeechT5ForTextToSpeech",
+            "SpeechT5HifiGan",
+            "SpeechT5Model",
+            "SpeechT5PreTrainedModel",
+        ]
+    )
     _import_structure["models.splinter"].extend(
         [
             "SPLINTER_PRETRAINED_MODEL_ARCHIVE_LIST",
@@ -3842,6 +3862,12 @@ if TYPE_CHECKING:
         Speech2Text2Processor,
         Speech2Text2Tokenizer,
     )
+    from .models.speecht5 import (
+        SPEECHT5_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        SPEECHT5_PRETRAINED_HIFIGAN_CONFIG_ARCHIVE_MAP,
+        SpeechT5Config,
+        SpeechT5HifiGanConfig,
+    )
     from .models.splinter import SPLINTER_PRETRAINED_CONFIG_ARCHIVE_MAP, SplinterConfig, SplinterTokenizer
     from .models.squeezebert import SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, SqueezeBertConfig, SqueezeBertTokenizer
     from .models.swin import SWIN_PRETRAINED_CONFIG_ARCHIVE_MAP, SwinConfig
@@ -4054,6 +4080,7 @@ if TYPE_CHECKING:
         from .models.reformer import ReformerTokenizer
         from .models.rembert import RemBertTokenizer
         from .models.speech_to_text import Speech2TextTokenizer
+        from .models.speecht5 import SpeechT5Tokenizer
         from .models.t5 import T5Tokenizer
         from .models.xglm import XGLMTokenizer
         from .models.xlm_prophetnet import XLMProphetNetTokenizer
@@ -4139,6 +4166,7 @@ if TYPE_CHECKING:
         from .models.audio_spectrogram_transformer import ASTFeatureExtractor
         from .models.mctct import MCTCTFeatureExtractor
         from .models.speech_to_text import Speech2TextFeatureExtractor
+        from .models.speecht5 import SpeechT5FeatureExtractor
 
     try:
         if not is_tensorflow_text_available():
@@ -4163,6 +4191,7 @@ if TYPE_CHECKING:
         from .utils.dummy_sentencepiece_and_speech_objects import *
     else:
         from .models.speech_to_text import Speech2TextProcessor
+        from .models.speecht5 import SpeechT5Processor
 
     try:
         if not is_vision_available():
@@ -5326,6 +5355,15 @@ if TYPE_CHECKING:
             Speech2TextPreTrainedModel,
         )
         from .models.speech_to_text_2 import Speech2Text2ForCausalLM, Speech2Text2PreTrainedModel
+        from .models.speecht5 import (
+            SPEECHT5_PRETRAINED_MODEL_ARCHIVE_LIST,
+            SpeechT5ForSpeechToSpeech,
+            SpeechT5ForSpeechToText,
+            SpeechT5ForTextToSpeech,
+            SpeechT5HifiGan,
+            SpeechT5Model,
+            SpeechT5PreTrainedModel,
+        )
         from .models.splinter import (
             SPLINTER_PRETRAINED_MODEL_ARCHIVE_LIST,
             SplinterForPreTraining,

src/transformers/modeling_outputs.py

@@ -1355,3 +1355,63 @@ class BaseModelOutputWithPoolingAndProjection(ModelOutput):
     hidden_states: Optional[Tuple[torch.FloatTensor]] = None
     attentions: Optional[Tuple[torch.FloatTensor]] = None
     projection_state: Optional[Tuple[torch.FloatTensor]] = None
+
+
+@dataclass
+class Seq2SeqSpectrogramOutput(ModelOutput):
+    """
+    Base class for sequence-to-sequence spectrogram outputs.
+
+    Args:
+        loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
+            Spectrogram generation loss.
+        spectrogram (`torch.FloatTensor` of shape `(batch_size, sequence_length, num_bins)`):
+            The predicted spectrogram.
+        past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
+            Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
+            `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape
+            `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`.
+
+            Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used (see `past_key_values` input) to speed up sequential decoding.
+        decoder_hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
+            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
+            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
+
+            Hidden-states of the decoder at the output of each layer plus the initial embedding outputs.
+        decoder_attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
+            sequence_length)`.
+
+            Attentions weights of the decoder, after the attention softmax, used to compute the weighted average in the
+            self-attention heads.
+        cross_attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
+            sequence_length)`.
+
+            Attentions weights of the decoder's cross-attention layer, after the attention softmax, used to compute the
+            weighted average in the cross-attention heads.
+        encoder_last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Sequence of hidden-states at the output of the last layer of the encoder of the model.
+        encoder_hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
+            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
+            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
+
+            Hidden-states of the encoder at the output of each layer plus the initial embedding outputs.
+        encoder_attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
+            sequence_length)`.
+
+            Attentions weights of the encoder, after the attention softmax, used to compute the weighted average in the
+            self-attention heads.
+    """
+
+    loss: Optional[torch.FloatTensor] = None
+    spectrogram: torch.FloatTensor = None
+    past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None
+    decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    cross_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    encoder_last_hidden_state: Optional[torch.FloatTensor] = None
+    encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    encoder_attentions: Optional[Tuple[torch.FloatTensor]] = None

src/transformers/models/__init__.py

@@ -155,6 +155,7 @@ from . import (
     speech_encoder_decoder,
     speech_to_text,
     speech_to_text_2,
+    speecht5,
     splinter,
     squeezebert,
     swin,

src/transformers/models/auto/configuration_auto.py

@@ -152,6 +152,7 @@ CONFIG_MAPPING_NAMES = OrderedDict(
         ("speech-encoder-decoder", "SpeechEncoderDecoderConfig"),
         ("speech_to_text", "Speech2TextConfig"),
         ("speech_to_text_2", "Speech2Text2Config"),
+        ("speecht5", "SpeechT5Config"),
         ("splinter", "SplinterConfig"),
         ("squeezebert", "SqueezeBertConfig"),
         ("swin", "SwinConfig"),
@@ -310,6 +311,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict(
         ("sew-d", "SEW_D_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("speech_to_text", "SPEECH_TO_TEXT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("speech_to_text_2", "SPEECH_TO_TEXT_2_PRETRAINED_CONFIG_ARCHIVE_MAP"),
+        ("speecht5", "SPEECHT5_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("splinter", "SPLINTER_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("squeezebert", "SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("swin", "SWIN_PRETRAINED_CONFIG_ARCHIVE_MAP"),
@@ -489,6 +491,7 @@ MODEL_NAMES_MAPPING = OrderedDict(
         ("speech-encoder-decoder", "Speech Encoder decoder"),
         ("speech_to_text", "Speech2Text"),
         ("speech_to_text_2", "Speech2Text2"),
+        ("speecht5", "SpeechT5"),
         ("splinter", "Splinter"),
         ("squeezebert", "SqueezeBERT"),
         ("swin", "Swin Transformer"),

src/transformers/models/auto/feature_extraction_auto.py

@@ -76,6 +76,7 @@ FEATURE_EXTRACTOR_MAPPING_NAMES = OrderedDict(
         ("sew", "Wav2Vec2FeatureExtractor"),
         ("sew-d", "Wav2Vec2FeatureExtractor"),
         ("speech_to_text", "Speech2TextFeatureExtractor"),
+        ("speecht5", "SpeechT5FeatureExtractor"),
         ("swin", "ViTFeatureExtractor"),
         ("swinv2", "ViTFeatureExtractor"),
         ("table-transformer", "DetrFeatureExtractor"),

src/transformers/models/auto/modeling_auto.py

@@ -148,6 +148,7 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("sew", "SEWModel"),
         ("sew-d", "SEWDModel"),
         ("speech_to_text", "Speech2TextModel"),
+        ("speecht5", "SpeechT5Model"),
         ("splinter", "SplinterModel"),
         ("squeezebert", "SqueezeBertModel"),
         ("swin", "SwinModel"),
@@ -591,6 +592,7 @@ MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES = OrderedDict(
     [
         ("speech-encoder-decoder", "SpeechEncoderDecoderModel"),
         ("speech_to_text", "Speech2TextForConditionalGeneration"),
+        ("speecht5", "SpeechT5ForSpeechToText"),
         ("whisper", "WhisperForConditionalGeneration"),
     ]
 )

src/transformers/models/auto/processing_auto.py

@@ -61,6 +61,7 @@ PROCESSOR_MAPPING_NAMES = OrderedDict(
         ("sew-d", "Wav2Vec2Processor"),
         ("speech_to_text", "Speech2TextProcessor"),
         ("speech_to_text_2", "Speech2Text2Processor"),
+        ("speecht5", "SpeechT5Processor"),
         ("trocr", "TrOCRProcessor"),
         ("unispeech", "Wav2Vec2Processor"),
         ("unispeech-sat", "Wav2Vec2Processor"),

src/transformers/models/auto/tokenization_auto.py

@@ -264,6 +264,7 @@ else:
             ("roformer", ("RoFormerTokenizer", "RoFormerTokenizerFast" if is_tokenizers_available() else None)),
             ("speech_to_text", ("Speech2TextTokenizer" if is_sentencepiece_available() else None, None)),
             ("speech_to_text_2", ("Speech2Text2Tokenizer", None)),
+            ("speecht5", ("SpeechT5Tokenizer" if is_sentencepiece_available() else None, None)),
             ("splinter", ("SplinterTokenizer", "SplinterTokenizerFast")),
             (
                 "squeezebert",

src/transformers/models/speecht5/__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 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
+
+from ...utils import (
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    is_sentencepiece_available,
+    is_speech_available,
+    is_torch_available,
+)
+
+
+_import_structure = {
+    "configuration_speecht5": [
+        "SPEECHT5_PRETRAINED_CONFIG_ARCHIVE_MAP",
+        "SPEECHT5_PRETRAINED_HIFIGAN_CONFIG_ARCHIVE_MAP",
+        "SpeechT5Config",
+        "SpeechT5HifiGanConfig",
+    ],
+}
+
+try:
+    if not is_sentencepiece_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["tokenization_speecht5"] = ["SpeechT5Tokenizer"]
+
+try:
+    if not is_speech_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["feature_extraction_speecht5"] = ["SpeechT5FeatureExtractor"]
+
+try:
+    if not (is_speech_available() and is_sentencepiece_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["processing_speecht5"] = ["SpeechT5Processor"]
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_speecht5"] = [
+        "SPEECHT5_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "SpeechT5ForSpeechToText",
+        "SpeechT5ForSpeechToSpeech",
+        "SpeechT5ForTextToSpeech",
+        "SpeechT5Model",
+        "SpeechT5PreTrainedModel",
+        "SpeechT5HifiGan",
+    ]
+
+if TYPE_CHECKING:
+    from .configuration_speecht5 import (
+        SPEECHT5_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        SPEECHT5_PRETRAINED_HIFIGAN_CONFIG_ARCHIVE_MAP,
+        SpeechT5Config,
+        SpeechT5HifiGanConfig,
+    )
+
+    try:
+        if not is_sentencepiece_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .tokenization_speecht5 import SpeechT5Tokenizer
+
+    try:
+        if not is_speech_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .feature_extraction_speecht5 import SpeechT5FeatureExtractor
+
+    try:
+        if not (is_speech_available() and is_sentencepiece_available()):
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .processing_speecht5 import SpeechT5Processor
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_speecht5 import (
+            SPEECHT5_PRETRAINED_MODEL_ARCHIVE_LIST,
+            SpeechT5ForSpeechToSpeech,
+            SpeechT5ForSpeechToText,
+            SpeechT5ForTextToSpeech,
+            SpeechT5HifiGan,
+            SpeechT5Model,
+            SpeechT5PreTrainedModel,
+        )
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)