Add modeling file

redirects should theoretically work, but still updating those repo references for clarity

.circleci/create_circleci_config.py

@@ -32,7 +32,8 @@ COMMON_ENV_VARIABLES = {
     "RUN_PT_TF_CROSS_TESTS": False,
     "RUN_PT_FLAX_CROSS_TESTS": False,
 }
-COMMON_PYTEST_OPTIONS = {"max-worker-restart": 0, "dist": "loadfile", "s": None}
+# Disable the use of {"s": None} as the output is way too long, causing the navigation on CircleCI impractical
+COMMON_PYTEST_OPTIONS = {"max-worker-restart": 0, "dist": "loadfile"}
 DEFAULT_DOCKER_IMAGE = [{"image": "cimg/python:3.8.12"}]
 
 
@@ -222,18 +223,27 @@ class CircleCIJob:
             # failure.
             test_command = f"({test_command}) || true"
         else:
-            test_command += " | tee tests_output.txt"
+            test_command += " || true"
         steps.append({"run": {"name": "Run tests", "command": test_command}})
 
+        check_test_command = f'if [ -s reports/{self.job_name}/failures_short.txt ]; '
+        check_test_command += 'then echo "Some test failed!"; echo ""; '
+        check_test_command += f'cat reports/{self.job_name}/failures_short.txt; '
+        check_test_command += 'echo ""; echo ""; '
+
+        py_command = f'import os; fp = open("reports/{self.job_name}/summary_short.txt"); failed = os.linesep.join([x for x in fp.read().split(os.linesep) if x.startswith(("FAILED ", "ERROR "))]); fp.close(); fp = open("summary_short.txt", "w"); fp.write(failed); fp.close()'
+        check_test_command += f"$(python3 -c '{py_command}'); "
+
+        check_test_command += f'cat summary_short.txt; echo ""; exit -1; '
+        check_test_command += f'elif [ -s reports/{self.job_name}/stats.txt ]; then echo "All tests pass!"; '
+
         # return code `124` means the previous (pytest run) step is timeout
         if self.name == "pr_documentation_tests":
-            checkout_doctest_command = 'if [ -s reports/tests_pr_documentation_tests/failures_short.txt ]; '
-            checkout_doctest_command += 'then echo "some test failed"; '
-            checkout_doctest_command += 'cat reports/tests_pr_documentation_tests/failures_short.txt; '
-            checkout_doctest_command += 'cat reports/tests_pr_documentation_tests/summary_short.txt; exit -1; '
-            checkout_doctest_command += 'elif [ -s reports/tests_pr_documentation_tests/stats.txt ]; then echo "All tests pass!"; '
-            checkout_doctest_command += 'elif [ -f 124.txt ]; then echo "doctest timeout!"; else echo "other fatal error)"; exit -1; fi;'
-            steps.append({"run": {"name": "Check doctest results", "command": checkout_doctest_command}})
+            check_test_command += 'elif [ -f 124.txt ]; then echo "doctest timeout!"; '
+
+        check_test_command += 'else echo "other fatal error"; echo ""; exit -1; fi;'
+
+        steps.append({"run": {"name": "Check test results", "command": check_test_command}})
 
         steps.append({"store_artifacts": {"path": "~/transformers/tests_output.txt"}})
         steps.append({"store_artifacts": {"path": "~/transformers/reports"}})
@@ -593,7 +603,7 @@ def create_circleci_config(folder=None):
                 job.tests_to_run = [f"examples/{framework}"]
             else:
                 job.tests_to_run = [f for f in example_tests.split(" ") if f.startswith(f"examples/{framework}")]
-            
+
             if len(job.tests_to_run) > 0:
                 jobs.append(job)
 

.github/workflows/update_metdata.yml

@@ -19,7 +19,7 @@ jobs:
       - name: Setup environment
         run: |
           pip install --upgrade pip
-          pip install datasets pandas
+          pip install datasets pandas==2.0.3
           pip install .[torch,tf,flax]
 
       - name: Update metadata

README.md

@@ -310,6 +310,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (from Alexa) released with the paper [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) by Adrian de Wynter and Daniel J. Perry.
 1. **[BridgeTower](https://huggingface.co/docs/transformers/model_doc/bridgetower)** (from Harbin Institute of Technology/Microsoft Research Asia/Intel Labs) released with the paper [BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning](https://arxiv.org/abs/2206.08657) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[BROS](https://huggingface.co/docs/transformers/model_doc/bros)** (from NAVER CLOVA) released with the paper [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539) by Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park.
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (from Google Research) released with the paper [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) by Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel.
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (from Inria/Facebook/Sorbonne) released with the paper [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
@@ -373,6 +374,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[GPTSAN-japanese](https://huggingface.co/docs/transformers/model_doc/gptsan-japanese)** released in the repository [tanreinama/GPTSAN](https://github.com/tanreinama/GPTSAN/blob/main/report/model.md) by Toshiyuki Sakamoto(tanreinama).
 1. **[Graphormer](https://huggingface.co/docs/transformers/model_doc/graphormer)** (from Microsoft) released with the paper [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234) by Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
+1. **[HerBERT](https://huggingface.co/docs/transformers/model_doc/herbert)** (from Allegro.pl, AGH University of Science and Technology) released with the paper [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf) by Piotr Rybak, Robert Mroczkowski, Janusz Tracz, Ireneusz Gawlik.
 1. **[Hubert](https://huggingface.co/docs/transformers/model_doc/hubert)** (from Facebook) released with the paper [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447) by Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed.
 1. **[I-BERT](https://huggingface.co/docs/transformers/model_doc/ibert)** (from Berkeley) released with the paper [I-BERT: Integer-only BERT Quantization](https://arxiv.org/abs/2101.01321) by Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer.
 1. **[IDEFICS](https://huggingface.co/docs/transformers/model_doc/idefics)** (from HuggingFace) released with the paper [OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents](https://huggingface.co/papers/2306.16527) by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sanh.
@@ -431,11 +433,12 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
 1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (from Google) released with the paper [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) by Jason Phang, Yao Zhao, and Peter J. Liu.
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
+1. **[Persimmon](https://huggingface.co/docs/transformers/main/model_doc/persimmon)** (from ADEPT) released in a [blog post](https://www.adept.ai/blog/persimmon-8b) by Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani.
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
 1. **[Pix2Struct](https://huggingface.co/docs/transformers/model_doc/pix2struct)** (from Google) released with the paper [Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding](https://arxiv.org/abs/2210.03347) by Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova.
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
 1. **[PoolFormer](https://huggingface.co/docs/transformers/model_doc/poolformer)** (from Sea AI Labs) released with the paper [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418) by Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng.
-1. **[Pop2Piano](https://huggingface.co/docs/transformers/main/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi and Kyogu Lee.
+1. **[Pop2Piano](https://huggingface.co/docs/transformers/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi and Kyogu Lee.
 1. **[ProphetNet](https://huggingface.co/docs/transformers/model_doc/prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
 1. **[PVT](https://huggingface.co/docs/transformers/model_doc/pvt)** (from Nanjing University, The University of Hong Kong etc.) released with the paper [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://arxiv.org/pdf/2102.12122.pdf) by Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao.
 1. **[QDQBert](https://huggingface.co/docs/transformers/model_doc/qdqbert)** (from NVIDIA) released with the paper [Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation](https://arxiv.org/abs/2004.09602) by Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius.
@@ -486,9 +489,10 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[ViT Hybrid](https://huggingface.co/docs/transformers/model_doc/vit_hybrid)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
-1. **[VitDet](https://huggingface.co/docs/transformers/main/model_doc/vitdet)** (from Meta AI) released with the paper [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) by Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.
+1. **[VitDet](https://huggingface.co/docs/transformers/model_doc/vitdet)** (from Meta AI) released with the paper [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) by Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
 1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (from Meta AI) released with the paper [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas.
+1. **[VITS](https://huggingface.co/docs/transformers/model_doc/vits)** (from Kakao Enterprise) released with the paper [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103) by Jaehyeon Kim, Jungil Kong, Juhee Son.
 1. **[ViViT](https://huggingface.co/docs/transformers/model_doc/vivit)** (from Google Research) released with the paper [ViViT: A Video Vision Transformer](https://arxiv.org/abs/2103.15691) by Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (from Facebook AI) released 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, Sravya Popuri, Dmytro Okhonko, Juan Pino.

README_es.md

@@ -287,6 +287,7 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt
 1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (from Alexa) released with the paper [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) by Adrian de Wynter and Daniel J. Perry.
 1. **[BridgeTower](https://huggingface.co/docs/transformers/model_doc/bridgetower)** (from Harbin Institute of Technology/Microsoft Research Asia/Intel Labs) released with the paper [BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning](https://arxiv.org/abs/2206.08657) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[BROS](https://huggingface.co/docs/transformers/model_doc/bros)** (from NAVER CLOVA) released with the paper [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539) by Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park.
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (from Google Research) released with the paper [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) by Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel.
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (from Inria/Facebook/Sorbonne) released with the paper [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
@@ -350,6 +351,7 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt
 1. **[GPTSAN-japanese](https://huggingface.co/docs/transformers/model_doc/gptsan-japanese)** released in the repository [tanreinama/GPTSAN](https://github.com/tanreinama/GPTSAN/blob/main/report/model.md) by Toshiyuki Sakamoto(tanreinama).
 1. **[Graphormer](https://huggingface.co/docs/transformers/model_doc/graphormer)** (from Microsoft) released with the paper [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234) by Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
+1. **[HerBERT](https://huggingface.co/docs/transformers/model_doc/herbert)** (from Allegro.pl, AGH University of Science and Technology) released with the paper [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf) by Piotr Rybak, Robert Mroczkowski, Janusz Tracz, Ireneusz Gawlik.
 1. **[Hubert](https://huggingface.co/docs/transformers/model_doc/hubert)** (from Facebook) released with the paper [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447) by Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed.
 1. **[I-BERT](https://huggingface.co/docs/transformers/model_doc/ibert)** (from Berkeley) released with the paper [I-BERT: Integer-only BERT Quantization](https://arxiv.org/abs/2101.01321) by Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer.
 1. **[IDEFICS](https://huggingface.co/docs/transformers/model_doc/idefics)** (from HuggingFace) released with the paper [OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents](https://huggingface.co/papers/2306.16527) by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sanh. 
@@ -408,11 +410,12 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
 1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (from Google) released with the paper [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) by Jason Phang, Yao Zhao, and Peter J. Liu.
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
+1. **[Persimmon](https://huggingface.co/docs/transformers/main/model_doc/persimmon)** (from ADEPT) released with the paper [blog post](https://www.adept.ai/blog/persimmon-8b) by Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani.
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
 1. **[Pix2Struct](https://huggingface.co/docs/transformers/model_doc/pix2struct)** (from Google) released with the paper [Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding](https://arxiv.org/abs/2210.03347) by Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova.
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
 1. **[PoolFormer](https://huggingface.co/docs/transformers/model_doc/poolformer)** (from Sea AI Labs) released with the paper [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418) by Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng.
-1. **[Pop2Piano](https://huggingface.co/docs/transformers/main/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
+1. **[Pop2Piano](https://huggingface.co/docs/transformers/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
 1. **[ProphetNet](https://huggingface.co/docs/transformers/model_doc/prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
 1. **[PVT](https://huggingface.co/docs/transformers/model_doc/pvt)** (from Nanjing University, The University of Hong Kong etc.) released with the paper [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://arxiv.org/pdf/2102.12122.pdf) by Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao.
 1. **[QDQBert](https://huggingface.co/docs/transformers/model_doc/qdqbert)** (from NVIDIA) released with the paper [Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation](https://arxiv.org/abs/2004.09602) by Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius.
@@ -463,9 +466,10 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[ViT Hybrid](https://huggingface.co/docs/transformers/model_doc/vit_hybrid)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
-1. **[VitDet](https://huggingface.co/docs/transformers/main/model_doc/vitdet)** (from Meta AI) released with the paper [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) by Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.
+1. **[VitDet](https://huggingface.co/docs/transformers/model_doc/vitdet)** (from Meta AI) released with the paper [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) by Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
 1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (from Meta AI) released with the paper [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas.
+1. **[VITS](https://huggingface.co/docs/transformers/model_doc/vits)** (from Kakao Enterprise) released with the paper [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103) by Jaehyeon Kim, Jungil Kong, Juhee Son.
 1. **[ViViT](https://huggingface.co/docs/transformers/model_doc/vivit)** (from Google Research) released with the paper [ViViT: A Video Vision Transformer](https://arxiv.org/abs/2103.15691) by Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (from Facebook AI) released 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, Sravya Popuri, Dmytro Okhonko, Juan Pino.

README_hd.md

@@ -259,6 +259,7 @@ conda install -c huggingface transformers
 1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigSicence Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (एलेक्सा से) कागज के साथ [बीईआरटी के लिए ऑप्टिमल सबआर्किटेक्चर एक्सट्रैक्शन](https://arxiv.org/abs/ 2010.10499) एड्रियन डी विंटर और डैनियल जे पेरी द्वारा।
 1. **[BridgeTower](https://huggingface.co/docs/transformers/model_doc/bridgetower)** (हरबिन इंस्टिट्यूट ऑफ़ टेक्नोलॉजी/माइक्रोसॉफ्ट रिसर्च एशिया/इंटेल लैब्स से) कागज के साथ [ब्रिजटॉवर: विजन-लैंग्वेज रिप्रेजेंटेशन लर्निंग में एनकोडर्स के बीच ब्रिज बनाना](<https://arxiv.org/abs/2206.08657>) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[BROS](https://huggingface.co/docs/transformers/model_doc/bros)** (NAVER CLOVA से) Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park. द्वाराअनुसंधान पत्र [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539) के साथ जारी किया गया
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (Google अनुसंधान से) साथ में कागज [ByT5: पूर्व-प्रशिक्षित बाइट-टू-बाइट मॉडल के साथ एक टोकन-मुक्त भविष्य की ओर] (https://arxiv.org/abs/2105.13626) Linting Xue, Aditya Barua, Noah Constant, रामी अल-रफू, शरण नारंग, मिहिर काले, एडम रॉबर्ट्स, कॉलिन रैफेल द्वारा पोस्ट किया गया।
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (इनरिया/फेसबुक/सोरबोन से) साथ में कागज [CamemBERT: एक टेस्टी फ्रेंच लैंग्वेज मॉडल](https:// arxiv.org/abs/1911.03894) लुई मार्टिन*, बेंजामिन मुलर*, पेड्रो जेवियर ऑर्टिज़ सुआरेज़*, योआन ड्यूपॉन्ट, लॉरेंट रोमरी, एरिक विलेमोन्टे डे ला क्लर्जरी, जैमे सेडाह और बेनोइट सगोट द्वारा।
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (Google रिसर्च से) साथ में दिया गया पेपर [कैनाइन: प्री-ट्रेनिंग ए एफिशिएंट टोकनाइजेशन-फ्री एनकोडर फॉर लैंग्वेज रिप्रेजेंटेशन]( https://arxiv.org/abs/2103.06874) जोनाथन एच क्लार्क, डैन गैरेट, यूलिया टर्क, जॉन विएटिंग द्वारा।
@@ -322,6 +323,7 @@ conda install -c huggingface transformers
 1. **[GPTSAN-japanese](https://huggingface.co/docs/transformers/model_doc/gptsan-japanese)** released in the repository [tanreinama/GPTSAN](https://github.com/tanreinama/GPTSAN/blob/main/report/model.md) by Toshiyuki Sakamoto(tanreinama).
 1. **[Graphormer](https://huggingface.co/docs/transformers/model_doc/graphormer)** (from Microsoft) released with the paper [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234) by Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (UCSD, NVIDIA से) साथ में कागज [GroupViT: टेक्स्ट सुपरविजन से सिमेंटिक सेगमेंटेशन इमर्जेस](https://arxiv .org/abs/2202.11094) जियारुई जू, शालिनी डी मेलो, सिफ़ी लियू, वोनमिन बायन, थॉमस ब्रेउएल, जान कौट्ज़, ज़ियाओलोंग वांग द्वारा।
+1. **[HerBERT](https://huggingface.co/docs/transformers/model_doc/herbert)** (Allegro.pl, AGH University of Science and Technology से) Piotr Rybak, Robert Mroczkowski, Janusz Tracz, Ireneusz Gawlik. द्वाराअनुसंधान पत्र [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf) के साथ जारी किया गया
 1. **[Hubert](https://huggingface.co/docs/transformers/model_doc/hubert)** (फेसबुक से) साथ में पेपर [ह्यूबर्ट: सेल्फ सुपरवाइज्ड स्पीच रिप्रेजेंटेशन लर्निंग बाय मास्क्ड प्रेडिक्शन ऑफ हिडन यूनिट्स](https ://arxiv.org/abs/2106.07447) वेई-निंग सू, बेंजामिन बोल्टे, याओ-हंग ह्यूबर्ट त्साई, कुशाल लखोटिया, रुस्लान सालाखुतदीनोव, अब्देलरहमान मोहम्मद द्वारा।
 1. **[I-BERT](https://huggingface.co/docs/transformers/model_doc/ibert)** (बर्कले से) साथ में कागज [I-BERT: Integer-only BERT Quantization](https:// arxiv.org/abs/2101.01321) सेहून किम, अमीर घोलमी, ज़ेवेई याओ, माइकल डब्ल्यू महोनी, कर्ट केटज़र द्वारा।
 1. **[IDEFICS](https://huggingface.co/docs/transformers/model_doc/idefics)** (from HuggingFace) released with the paper [OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents](https://huggingface.co/papers/2306.16527) by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sanh. 
@@ -380,11 +382,12 @@ conda install -c huggingface transformers
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
 1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (Google की ओर से) साथ में दिया गया पेपर [लंबे इनपुट सारांश के लिए ट्रांसफ़ॉर्मरों को बेहतर तरीके से एक्सटेंड करना](https://arxiv .org/abs/2208.04347) जेसन फांग, याओ झाओ, पीटर जे लियू द्वारा।
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (दीपमाइंड से) साथ में पेपर [पर्सीवर आईओ: संरचित इनपुट और आउटपुट के लिए एक सामान्य वास्तुकला] (https://arxiv.org/abs/2107.14795) एंड्रयू जेगल, सेबेस्टियन बोरग्यूड, जीन-बैप्टिस्ट अलायराक, कार्ल डोर्श, कैटलिन इओनेस्कु, डेविड द्वारा डिंग, स्कंद कोप्पुला, डैनियल ज़ोरान, एंड्रयू ब्रॉक, इवान शेलहैमर, ओलिवियर हेनाफ, मैथ्यू एम। बोट्विनिक, एंड्रयू ज़िसरमैन, ओरिओल विनियल्स, जोआओ कैरेरा द्वारा पोस्ट किया गया।
+1. **[Persimmon](https://huggingface.co/docs/transformers/main/model_doc/persimmon)** (ADEPT से) Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani. द्वाराअनुसंधान पत्र [blog post](https://www.adept.ai/blog/persimmon-8b) के साथ जारी किया गया
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (VinAI Research से) कागज के साथ [PhoBERT: वियतनामी के लिए पूर्व-प्रशिक्षित भाषा मॉडल](https://www .aclweb.org/anthology/2020.findings-emnlp.92/) डैट क्वोक गुयेन और अन्ह तुआन गुयेन द्वारा पोस्ट किया गया।
 1. **[Pix2Struct](https://huggingface.co/docs/transformers/model_doc/pix2struct)** (Google से) Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova. द्वाराअनुसंधान पत्र [Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding](https://arxiv.org/abs/2210.03347) के साथ जारी किया गया
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (UCLA NLP से) साथ वाला पेपर [प्रोग्राम अंडरस्टैंडिंग एंड जेनरेशन के लिए यूनिफाइड प्री-ट्रेनिंग](https://arxiv .org/abs/2103.06333) वसी उद्दीन अहमद, सैकत चक्रवर्ती, बैशाखी रे, काई-वेई चांग द्वारा।
 1. **[PoolFormer](https://huggingface.co/docs/transformers/model_doc/poolformer)** (from Sea AI Labs) released with the paper [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418) by Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng.
-1. **[Pop2Piano](https://huggingface.co/docs/transformers/main/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
+1. **[Pop2Piano](https://huggingface.co/docs/transformers/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
 1. **[ProphetNet](https://huggingface.co/docs/transformers/model_doc/prophetnet)** (माइक्रोसॉफ्ट रिसर्च से) साथ में पेपर [ProphetNet: प्रेडिक्टिंग फ्यूचर एन-ग्राम फॉर सीक्वेंस-टू-सीक्वेंस प्री-ट्रेनिंग ](https://arxiv.org/abs/2001.04063) यू यान, वीज़ेन क्यूई, येयुन गोंग, दयाहेंग लियू, नान डुआन, जिउशेंग चेन, रुओफ़ेई झांग और मिंग झोउ द्वारा पोस्ट किया गया।
 1. **[PVT](https://huggingface.co/docs/transformers/model_doc/pvt)** (Nanjing University, The University of Hong Kong etc. से) Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao. द्वाराअनुसंधान पत्र [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://arxiv.org/pdf/2102.12122.pdf) के साथ जारी किया गया
 1. **[QDQBert](https://huggingface.co/docs/transformers/model_doc/qdqbert)** (NVIDIA से) साथ वाला पेपर [डीप लर्निंग इंफ़ेक्शन के लिए इंटीजर क्वांटिज़ेशन: प्रिंसिपल्स एंड एम्पिरिकल इवैल्यूएशन](https:// arxiv.org/abs/2004.09602) हाओ वू, पैट्रिक जुड, जिआओजी झांग, मिखाइल इसेव और पॉलियस माइकेविसियस द्वारा।
@@ -435,9 +438,10 @@ conda install -c huggingface transformers
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (गूगल एआई से) कागज के साथ [एक इमेज इज़ वर्थ 16x16 वर्ड्स: ट्रांसफॉर्मर्स फॉर इमेज रिकॉग्निशन एट स्केल](https://arxiv.org/abs/2010.11929) एलेक्सी डोसोवित्स्की, लुकास बेयर, अलेक्जेंडर कोलेसनिकोव, डिर्क वीसेनबोर्न, शियाओहुआ झाई, थॉमस अनटरथिनर, मुस्तफा देहघानी, मैथियास मिंडरर, जॉर्ज हेगोल्ड, सिल्वेन गेली, जैकब उस्ज़कोरेइट द्वारा हॉल्सबी द्वारा पोस्ट किया गया।
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (UCLA NLP से) साथ वाला पेपर [VisualBERT: A Simple and Performant Baseline for Vision and Language](https:/ /arxiv.org/pdf/1908.03557) लियुनियन हेरोल्ड ली, मार्क यात्स्कर, दा यिन, चो-जुई हसीह, काई-वेई चांग द्वारा।
 1. **[ViT Hybrid](https://huggingface.co/docs/transformers/model_doc/vit_hybrid)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
-1. **[VitDet](https://huggingface.co/docs/transformers/main/model_doc/vitdet)** (Meta AI से) Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He. द्वाराअनुसंधान पत्र [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) के साथ जारी किया गया
+1. **[VitDet](https://huggingface.co/docs/transformers/model_doc/vitdet)** (Meta AI से) Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He. द्वाराअनुसंधान पत्र [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) के साथ जारी किया गया
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (मेटा एआई से) साथ में कागज [मास्कड ऑटोएन्कोडर स्केलेबल विजन लर्नर्स हैं](https://arxiv.org/ एब्स/2111.06377) कैमिंग हे, ज़िनेली चेन, सेनिंग ज़ी, यांगहो ली, पिओट्र डॉलर, रॉस गिर्शिक द्वारा।
 1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (मेटा एआई से) साथ में कागज [लेबल-कुशल सीखने के लिए मास्क्ड स्याम देश के नेटवर्क](https://arxiv. org/abs/2204.07141) महमूद असरान, मथिल्डे कैरन, ईशान मिश्रा, पियोट्र बोजानोवस्की, फ्लोरियन बोर्डेस, पास्कल विंसेंट, आर्मंड जौलिन, माइकल रब्बत, निकोलस बल्लास द्वारा।
+1. **[VITS](https://huggingface.co/docs/transformers/model_doc/vits)** (Kakao Enterprise से) Jaehyeon Kim, Jungil Kong, Juhee Son. द्वाराअनुसंधान पत्र [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103) के साथ जारी किया गया
 1. **[ViViT](https://huggingface.co/docs/transformers/model_doc/vivit)** (from Google Research) released with the paper [ViViT: A Video Vision Transformer](https://arxiv.org/abs/2103.15691) by Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (फेसबुक एआई से) साथ में पेपर [wav2vec 2.0: ए फ्रेमवर्क फॉर सेल्फ-सुपरवाइज्ड लर्निंग ऑफ स्पीच रिप्रेजेंटेशन] (https://arxiv.org/abs/2006.11477) एलेक्सी बेवस्की, हेनरी झोउ, अब्देलरहमान मोहम्मद, माइकल औली द्वारा।
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (Facebook AI से) साथ वाला पेपर [FAIRSEQ S2T: FAIRSEQ के साथ फास्ट स्पीच-टू-टेक्स्ट मॉडलिंग ](https://arxiv.org/abs/2010.05171) चांगहान वांग, यूं तांग, जुताई मा, ऐनी वू, सरव्या पोपुरी, दिमित्रो ओखोनको, जुआन पिनो द्वारा पोस्ट किया गया।

README_ja.md

@@ -321,6 +321,7 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ
 1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (BigScience workshop から) [BigScience Workshop](https://bigscience.huggingface.co/) から公開されました.
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (Alexa から) Adrian de Wynter and Daniel J. Perry から公開された研究論文: [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499)
 1. **[BridgeTower](https://huggingface.co/docs/transformers/model_doc/bridgetower)** (Harbin Institute of Technology/Microsoft Research Asia/Intel Labs から) released with the paper [BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning](https://arxiv.org/abs/2206.08657) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[BROS](https://huggingface.co/docs/transformers/model_doc/bros)** (NAVER CLOVA から) Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park. から公開された研究論文 [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539)
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (Google Research から) Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel から公開された研究論文: [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626)
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (Inria/Facebook/Sorbonne から) Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot から公開された研究論文: [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894)
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (Google Research から) Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting から公開された研究論文: [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874)
@@ -384,6 +385,7 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ
 1. **[GPTSAN-japanese](https://huggingface.co/docs/transformers/model_doc/gptsan-japanese)** [tanreinama/GPTSAN](https://github.com/tanreinama/GPTSAN/blob/main/report/model.md) 坂本俊之(tanreinama)からリリースされました.
 1. **[Graphormer](https://huggingface.co/docs/transformers/model_doc/graphormer)** (Microsoft から) Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu から公開された研究論文: [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234).
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (UCSD, NVIDIA から) Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang から公開された研究論文: [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094)
+1. **[HerBERT](https://huggingface.co/docs/transformers/model_doc/herbert)** (Allegro.pl, AGH University of Science and Technology から) Piotr Rybak, Robert Mroczkowski, Janusz Tracz, Ireneusz Gawlik. から公開された研究論文 [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf)
 1. **[Hubert](https://huggingface.co/docs/transformers/model_doc/hubert)** (Facebook から) Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed から公開された研究論文: [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447)
 1. **[I-BERT](https://huggingface.co/docs/transformers/model_doc/ibert)** (Berkeley から) Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer から公開された研究論文: [I-BERT: Integer-only BERT Quantization](https://arxiv.org/abs/2101.01321)
 1. **[IDEFICS](https://huggingface.co/docs/transformers/model_doc/idefics)** (from HuggingFace) released with the paper [OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents](https://huggingface.co/papers/2306.16527) by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sanh. 
@@ -442,11 +444,12 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (Google から) Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu から公開された研究論文: [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777)
 1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (Google から) Jason Phang, Yao Zhao, and Peter J. Liu から公開された研究論文: [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347)
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (Deepmind から) Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira から公開された研究論文: [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795)
+1. **[Persimmon](https://huggingface.co/docs/transformers/main/model_doc/persimmon)** (ADEPT から) Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani. から公開された研究論文 [blog post](https://www.adept.ai/blog/persimmon-8b)
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (VinAI Research から) Dat Quoc Nguyen and Anh Tuan Nguyen から公開された研究論文: [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/)
 1. **[Pix2Struct](https://huggingface.co/docs/transformers/model_doc/pix2struct)** (Google から) Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova. から公開された研究論文 [Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding](https://arxiv.org/abs/2210.03347)
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (UCLA NLP から) Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang から公開された研究論文: [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333)
 1. **[PoolFormer](https://huggingface.co/docs/transformers/model_doc/poolformer)** (Sea AI Labs から) Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng から公開された研究論文: [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418)
-1. **[Pop2Piano](https://huggingface.co/docs/transformers/main/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
+1. **[Pop2Piano](https://huggingface.co/docs/transformers/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
 1. **[ProphetNet](https://huggingface.co/docs/transformers/model_doc/prophetnet)** (Microsoft Research から) Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou から公開された研究論文: [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063)
 1. **[PVT](https://huggingface.co/docs/transformers/model_doc/pvt)** (Nanjing University, The University of Hong Kong etc. から) Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao. から公開された研究論文 [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://arxiv.org/pdf/2102.12122.pdf)
 1. **[QDQBert](https://huggingface.co/docs/transformers/model_doc/qdqbert)** (NVIDIA から) Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius から公開された研究論文: [Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation](https://arxiv.org/abs/2004.09602)
@@ -497,9 +500,10 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (Google AI から) Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby から公開された研究論文: [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929)
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (UCLA NLP から) Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang から公開された研究論文: [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557)
 1. **[ViT Hybrid](https://huggingface.co/docs/transformers/model_doc/vit_hybrid)** (Google AI から) Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby から公開された研究論文: [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929)
-1. **[VitDet](https://huggingface.co/docs/transformers/main/model_doc/vitdet)** (Meta AI から) Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He. から公開された研究論文 [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527)
+1. **[VitDet](https://huggingface.co/docs/transformers/model_doc/vitdet)** (Meta AI から) Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He. から公開された研究論文 [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527)
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (Meta AI から) Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick から公開された研究論文: [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377)
 1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (Meta AI から) Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas から公開された研究論文: [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141)
+1. **[VITS](https://huggingface.co/docs/transformers/model_doc/vits)** (Kakao Enterprise から) Jaehyeon Kim, Jungil Kong, Juhee Son. から公開された研究論文 [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103)
 1. **[ViViT](https://huggingface.co/docs/transformers/model_doc/vivit)** (from Google Research) released with the paper [ViViT: A Video Vision Transformer](https://arxiv.org/abs/2103.15691) by Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (Facebook AI から) Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli から公開された研究論文: [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477)
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (Facebook AI から) Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, 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)

README_ko.md

@@ -236,6 +236,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (Alexa 에서) Adrian de Wynter and Daniel J. Perry 의 [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) 논문과 함께 발표했습니다.
 1. **[BridgeTower](https://huggingface.co/docs/transformers/model_doc/bridgetower)** (from Harbin Institute of Technology/Microsoft Research Asia/Intel Labs) released with the paper [BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning](https://arxiv.org/abs/2206.08657) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[BROS](https://huggingface.co/docs/transformers/model_doc/bros)** (NAVER CLOVA 에서 제공)은 Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park.의 [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539)논문과 함께 발표했습니다.
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (Google Research 에서) Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel 의 [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) 논문과 함께 발표했습니다.
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (Inria/Facebook/Sorbonne 에서) Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot 의 [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) 논문과 함께 발표했습니다.
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (Google Research 에서) Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting 의 [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) 논문과 함께 발표했습니다.
@@ -299,6 +300,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[GPTSAN-japanese](https://huggingface.co/docs/transformers/model_doc/gptsan-japanese)** released in the repository [tanreinama/GPTSAN](https://github.com/tanreinama/GPTSAN/blob/main/report/model.md) by Toshiyuki Sakamoto(tanreinama).
 1. **[Graphormer](https://huggingface.co/docs/transformers/model_doc/graphormer)** (from Microsoft) Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu  의 [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234)  논문과 함께 발표했습니다.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (UCSD, NVIDIA 에서) Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang 의 [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) 논문과 함께 발표했습니다.
+1. **[HerBERT](https://huggingface.co/docs/transformers/model_doc/herbert)** (Allegro.pl, AGH University of Science and Technology 에서 제공)은 Piotr Rybak, Robert Mroczkowski, Janusz Tracz, Ireneusz Gawlik.의 [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf)논문과 함께 발표했습니다.
 1. **[Hubert](https://huggingface.co/docs/transformers/model_doc/hubert)** (Facebook 에서) Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed 의 [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447) 논문과 함께 발표했습니다.
 1. **[I-BERT](https://huggingface.co/docs/transformers/model_doc/ibert)** (Berkeley 에서) Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer 의 [I-BERT: Integer-only BERT Quantization](https://arxiv.org/abs/2101.01321) 논문과 함께 발표했습니다.
 1. **[IDEFICS](https://huggingface.co/docs/transformers/model_doc/idefics)** (from HuggingFace) released with the paper [OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents](https://huggingface.co/papers/2306.16527) by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sanh. 
@@ -357,11 +359,12 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (Google 에서) Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu 의 [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) 논문과 함께 발표했습니다.
 1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (Google 에서) Jason Phang, Yao Zhao, Peter J. Liu 의 [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) 논문과 함께 발표했습니다.
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (Deepmind 에서) Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira 의 [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) 논문과 함께 발표했습니다.
+1. **[Persimmon](https://huggingface.co/docs/transformers/main/model_doc/persimmon)** (ADEPT 에서 제공)은 Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani.의 [blog post](https://www.adept.ai/blog/persimmon-8b)논문과 함께 발표했습니다.
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (VinAI Research 에서) Dat Quoc Nguyen and Anh Tuan Nguyen 의 [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) 논문과 함께 발표했습니다.
 1. **[Pix2Struct](https://huggingface.co/docs/transformers/model_doc/pix2struct)** (Google 에서 제공)은 Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova.의 [Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding](https://arxiv.org/abs/2210.03347)논문과 함께 발표했습니다.
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (UCLA NLP 에서) Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang 의 [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) 논문과 함께 발표했습니다.
 1. **[PoolFormer](https://huggingface.co/docs/transformers/model_doc/poolformer)** (Sea AI Labs 에서) Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng 의 [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418) 논문과 함께 발표했습니다.
-1. **[Pop2Piano](https://huggingface.co/docs/transformers/main/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
+1. **[Pop2Piano](https://huggingface.co/docs/transformers/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
 1. **[ProphetNet](https://huggingface.co/docs/transformers/model_doc/prophetnet)** (Microsoft Research 에서) Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou 의 [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) 논문과 함께 발표했습니다.
 1. **[PVT](https://huggingface.co/docs/transformers/model_doc/pvt)** (Nanjing University, The University of Hong Kong etc. 에서 제공)은 Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao.의 [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://arxiv.org/pdf/2102.12122.pdf)논문과 함께 발표했습니다.
 1. **[QDQBert](https://huggingface.co/docs/transformers/model_doc/qdqbert)** (NVIDIA 에서) Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius 의 [Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation](https://arxiv.org/abs/2004.09602) 논문과 함께 발표했습니다.
@@ -412,9 +415,10 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (Google AI 에서) Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby 의 [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) 논문과 함께 발표했습니다.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (UCLA NLP 에서) Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang 의 [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) 논문과 함께 발표했습니다.
 1. **[ViT Hybrid](https://huggingface.co/docs/transformers/model_doc/vit_hybrid)** (Google AI 에서) Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby 의 [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) 논문과 함께 발표했습니다.
-1. **[VitDet](https://huggingface.co/docs/transformers/main/model_doc/vitdet)** (Meta AI 에서 제공)은 Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.의 [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527)논문과 함께 발표했습니다.
+1. **[VitDet](https://huggingface.co/docs/transformers/model_doc/vitdet)** (Meta AI 에서 제공)은 Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.의 [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527)논문과 함께 발표했습니다.
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (Meta AI 에서) Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick 의 [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) 논문과 함께 발표했습니다.
 1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (Meta AI 에서) Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas 의 [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) 논문과 함께 발표했습니다.
+1. **[VITS](https://huggingface.co/docs/transformers/model_doc/vits)** (Kakao Enterprise 에서 제공)은 Jaehyeon Kim, Jungil Kong, Juhee Son.의 [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103)논문과 함께 발표했습니다.
 1. **[ViViT](https://huggingface.co/docs/transformers/model_doc/vivit)** (from Google Research) released with the paper [ViViT: A Video Vision Transformer](https://arxiv.org/abs/2103.15691) by Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (Facebook AI 에서) Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli 의 [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) 논문과 함께 발표했습니다.
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (Facebook AI 에서) Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino 의 [FAIRSEQ S2T: Fast Speech-to-Text Modeling with FAIRSEQ](https://arxiv.org/abs/2010.05171) 논문과 함께 발표했습니다.
@@ -433,7 +437,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[XLS-R](https://huggingface.co/docs/transformers/model_doc/xls_r)** (Facebook AI 에서) Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, Kritika Singh, Patrick von Platen, Yatharth Saraf, Juan Pino, Alexei Baevski, Alexis Conneau, Michael Auli 의 [XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale](https://arxiv.org/abs/2111.09296) 논문과 함께 발표했습니다.
 1. **[XLSR-Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/xlsr_wav2vec2)** (Facebook AI 에서) Alexis Conneau, Alexei Baevski, Ronan Collobert, Abdelrahman Mohamed, Michael Auli 의 [Unsupervised Cross-Lingual Representation Learning For Speech Recognition](https://arxiv.org/abs/2006.13979) 논문과 함께 발표했습니다.
 1. **[YOLOS](https://huggingface.co/docs/transformers/model_doc/yolos)** (Huazhong University of Science & Technology 에서) Yuxin Fang, Bencheng Liao, Xinggang Wang, Jiemin Fang, Jiyang Qi, Rui Wu, Jianwei Niu, Wenyu Liu 의 [You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection](https://arxiv.org/abs/2106.00666) 논문과 함께 발표했습니다.
-1. **[YOSO](https://huggingface.co/docs/transformers/model_doc/yoso)** (the University of Wisconsin - Madison 에서) Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh 의 [You Only Sample (Almost) 논문과 함께 발표했습니다.
+1. **[YOSO](https://huggingface.co/docs/transformers/model_doc/yoso)** (the University of Wisconsin - Madison 에서) Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh 의 [You Only Sample (Almost) Once: Linear Cost Self-Attention Via Bernoulli Sampling](https://arxiv.org/abs/2111.09714) 논문과 함께 발표했습니다.
 1. 새로운 모델을 올리고 싶나요? 우리가 **상세한 가이드와 템플릿** 으로 새로운 모델을 올리도록 도와드릴게요. 가이드와 템플릿은 이 저장소의 [`templates`](./templates) 폴더에서 확인하실 수 있습니다. [컨트리뷰션 가이드라인](./CONTRIBUTING.md)을 꼭 확인해주시고, PR을 올리기 전에 메인테이너에게 연락하거나 이슈를 오픈해 피드백을 받으시길 바랍니다.
 
 각 모델이 Flax, PyTorch, TensorFlow으로 구현되었는지 또는 🤗 Tokenizers 라이브러리가 지원하는 토크나이저를 사용하는지 확인하려면, [이 표](https://huggingface.co/docs/transformers/index#supported-frameworks)를 확인하세요.

README_zh-hans.md

@@ -260,6 +260,7 @@ conda install -c huggingface transformers
 1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (来自 Alexa) 伴随论文 [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) 由 Adrian de Wynter and Daniel J. Perry 发布。
 1. **[BridgeTower](https://huggingface.co/docs/transformers/model_doc/bridgetower)** (from Harbin Institute of Technology/Microsoft Research Asia/Intel Labs) released with the paper [BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning](https://arxiv.org/abs/2206.08657) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[BROS](https://huggingface.co/docs/transformers/model_doc/bros)** (来自 NAVER CLOVA) 伴随论文 [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539) 由 Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park 发布。
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (来自 Google Research) 伴随论文 [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) 由 Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel 发布。
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (来自 Inria/Facebook/Sorbonne) 伴随论文 [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) 由 Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot 发布。
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (来自 Google Research) 伴随论文 [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) 由 Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting 发布。
@@ -323,6 +324,7 @@ conda install -c huggingface transformers
 1. **[GPTSAN-japanese](https://huggingface.co/docs/transformers/model_doc/gptsan-japanese)** released in the repository [tanreinama/GPTSAN](https://github.com/tanreinama/GPTSAN/blob/main/report/model.md) by 坂本俊之(tanreinama).
 1. **[Graphormer](https://huggingface.co/docs/transformers/model_doc/graphormer)** (from Microsoft) released with the paper [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234) by Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (来自 UCSD, NVIDIA) 伴随论文 [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) 由 Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang 发布。
+1. **[HerBERT](https://huggingface.co/docs/transformers/model_doc/herbert)** (来自 Allegro.pl, AGH University of Science and Technology) 伴随论文 [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf) 由 Piotr Rybak, Robert Mroczkowski, Janusz Tracz, Ireneusz Gawlik 发布。
 1. **[Hubert](https://huggingface.co/docs/transformers/model_doc/hubert)** (来自 Facebook) 伴随论文 [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447) 由 Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed 发布。
 1. **[I-BERT](https://huggingface.co/docs/transformers/model_doc/ibert)** (来自 Berkeley) 伴随论文 [I-BERT: Integer-only BERT Quantization](https://arxiv.org/abs/2101.01321) 由 Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer 发布。
 1. **[IDEFICS](https://huggingface.co/docs/transformers/model_doc/idefics)** (from HuggingFace) released with the paper [OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents](https://huggingface.co/papers/2306.16527) by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sanh. 
@@ -381,11 +383,12 @@ conda install -c huggingface transformers
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (来自 Google) 伴随论文 [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) 由 Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu 发布。
 1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (来自 Google) 伴随论文 [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) 由 Jason Phang, Yao Zhao, Peter J. Liu 发布。
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (来自 Deepmind) 伴随论文 [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) 由 Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira 发布。
+1. **[Persimmon](https://huggingface.co/docs/transformers/main/model_doc/persimmon)** (来自 ADEPT) 伴随论文 [blog post](https://www.adept.ai/blog/persimmon-8b) 由 Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani 发布。
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (来自 VinAI Research) 伴随论文 [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) 由 Dat Quoc Nguyen and Anh Tuan Nguyen 发布。
 1. **[Pix2Struct](https://huggingface.co/docs/transformers/model_doc/pix2struct)** (来自 Google) 伴随论文 [Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding](https://arxiv.org/abs/2210.03347) 由 Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova 发布。
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (来自 UCLA NLP) 伴随论文 [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) 由 Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang 发布。
 1. **[PoolFormer](https://huggingface.co/docs/transformers/model_doc/poolformer)** (来自 Sea AI Labs) 伴随论文 [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418) 由 Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng 发布。
-1. **[Pop2Piano](https://huggingface.co/docs/transformers/main/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
+1. **[Pop2Piano](https://huggingface.co/docs/transformers/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
 1. **[ProphetNet](https://huggingface.co/docs/transformers/model_doc/prophetnet)** (来自 Microsoft Research) 伴随论文 [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) 由 Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou 发布。
 1. **[PVT](https://huggingface.co/docs/transformers/model_doc/pvt)** (来自 Nanjing University, The University of Hong Kong etc.) 伴随论文 [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://arxiv.org/pdf/2102.12122.pdf) 由 Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao 发布。
 1. **[QDQBert](https://huggingface.co/docs/transformers/model_doc/qdqbert)** (来自 NVIDIA) 伴随论文 [Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation](https://arxiv.org/abs/2004.09602) 由 Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius 发布。
@@ -436,9 +439,10 @@ conda install -c huggingface transformers
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (来自 Google AI) 伴随论文 [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) 由 Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby 发布。
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (来自 UCLA NLP) 伴随论文 [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) 由 Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang 发布。
 1. **[ViT Hybrid](https://huggingface.co/docs/transformers/model_doc/vit_hybrid)** (来自 Google AI) 伴随论文 [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) 由 Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby 发布。
-1. **[VitDet](https://huggingface.co/docs/transformers/main/model_doc/vitdet)** (来自 Meta AI) 伴随论文 [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) 由 Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He 发布。
+1. **[VitDet](https://huggingface.co/docs/transformers/model_doc/vitdet)** (来自 Meta AI) 伴随论文 [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) 由 Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He 发布。
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (来自 Meta AI) 伴随论文 [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) 由 Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick 发布。
 1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (来自 Meta AI) 伴随论文 [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas 发布.
+1. **[VITS](https://huggingface.co/docs/transformers/model_doc/vits)** (来自 Kakao Enterprise) 伴随论文 [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103) 由 Jaehyeon Kim, Jungil Kong, Juhee Son 发布。
 1. **[ViViT](https://huggingface.co/docs/transformers/model_doc/vivit)** (来自 Google Research) released with the paper [ViViT: A Video Vision Transformer](https://arxiv.org/abs/2103.15691) 由 Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (来自 Facebook AI) 伴随论文 [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) 由 Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli 发布。
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (来自 Facebook AI) 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, Sravya Popuri, Dmytro Okhonko, Juan Pino 发布。
@@ -457,7 +461,7 @@ conda install -c huggingface transformers
 1. **[XLS-R](https://huggingface.co/docs/transformers/model_doc/xls_r)** (来自 Facebook AI) 伴随论文 [XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale](https://arxiv.org/abs/2111.09296) 由 Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, Kritika Singh, Patrick von Platen, Yatharth Saraf, Juan Pino, Alexei Baevski, Alexis Conneau, Michael Auli 发布。
 1. **[XLSR-Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/xlsr_wav2vec2)** (来自 Facebook AI) 伴随论文 [Unsupervised Cross-Lingual Representation Learning For Speech Recognition](https://arxiv.org/abs/2006.13979) 由 Alexis Conneau, Alexei Baevski, Ronan Collobert, Abdelrahman Mohamed, Michael Auli 发布。
 1. **[YOLOS](https://huggingface.co/docs/transformers/model_doc/yolos)** (来自 Huazhong University of Science & Technology) 伴随论文 [You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection](https://arxiv.org/abs/2106.00666) 由 Yuxin Fang, Bencheng Liao, Xinggang Wang, Jiemin Fang, Jiyang Qi, Rui Wu, Jianwei Niu, Wenyu Liu 发布。
-1. **[YOSO](https://huggingface.co/docs/transformers/model_doc/yoso)** (来自 the University of Wisconsin - Madison) 伴随论文 [You Only Sample (Almost) 由 Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh 发布。
+1. **[YOSO](https://huggingface.co/docs/transformers/model_doc/yoso)** (来自 the University of Wisconsin - Madison) 伴随论文 [You Only Sample (Almost) Once: Linear Cost Self-Attention Via Bernoulli Sampling](https://arxiv.org/abs/2111.09714) 由 Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh 发布。
 1. 想要贡献新的模型？我们这里有一份**详细指引和模板**来引导你添加新的模型。你可以在 [`templates`](./templates) 目录中找到他们。记得查看 [贡献指南](./CONTRIBUTING.md) 并在开始写 PR 前联系维护人员或开一个新的 issue 来获得反馈。
 
 要检查某个模型是否已有 Flax、PyTorch 或 TensorFlow 的实现，或其是否在 🤗 Tokenizers 库中有对应词符化器（tokenizer），敬请参阅[此表](https://huggingface.co/docs/transformers/index#supported-frameworks)。

README_zh-hant.md

@@ -272,6 +272,7 @@ conda install -c huggingface transformers
 1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (from Alexa) released with the paper [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) by Adrian de Wynter and Daniel J. Perry.
 1. **[BridgeTower](https://huggingface.co/docs/transformers/model_doc/bridgetower)** (from Harbin Institute of Technology/Microsoft Research Asia/Intel Labs) released with the paper [BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning](https://arxiv.org/abs/2206.08657) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[BROS](https://huggingface.co/docs/transformers/model_doc/bros)** (from NAVER CLOVA) released with the paper [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539) by Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park.
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (from Google Research) released with the paper [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) by Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel.
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (from Inria/Facebook/Sorbonne) released with the paper [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
@@ -335,6 +336,7 @@ conda install -c huggingface transformers
 1. **[GPTSAN-japanese](https://huggingface.co/docs/transformers/model_doc/gptsan-japanese)** released in the repository [tanreinama/GPTSAN](https://github.com/tanreinama/GPTSAN/blob/main/report/model.md) by 坂本俊之(tanreinama).
 1. **[Graphormer](https://huggingface.co/docs/transformers/model_doc/graphormer)** (from Microsoft) released with the paper [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234) by Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
+1. **[HerBERT](https://huggingface.co/docs/transformers/model_doc/herbert)** (from Allegro.pl, AGH University of Science and Technology) released with the paper [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf) by Piotr Rybak, Robert Mroczkowski, Janusz Tracz, Ireneusz Gawlik.
 1. **[Hubert](https://huggingface.co/docs/transformers/model_doc/hubert)** (from Facebook) released with the paper [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447) by Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed.
 1. **[I-BERT](https://huggingface.co/docs/transformers/model_doc/ibert)** (from Berkeley) released with the paper [I-BERT: Integer-only BERT Quantization](https://arxiv.org/abs/2101.01321) by Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer.
 1. **[IDEFICS](https://huggingface.co/docs/transformers/model_doc/idefics)** (from HuggingFace) released with the paper [OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents](https://huggingface.co/papers/2306.16527) by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sanh. 
@@ -393,11 +395,12 @@ conda install -c huggingface transformers
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
 1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (from Google) released with the paper [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) by Jason Phang, Yao Zhao, Peter J. Liu.
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
+1. **[Persimmon](https://huggingface.co/docs/transformers/main/model_doc/persimmon)** (from ADEPT) released with the paper [blog post](https://www.adept.ai/blog/persimmon-8b) by Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani.
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
 1. **[Pix2Struct](https://huggingface.co/docs/transformers/model_doc/pix2struct)** (from Google) released with the paper [Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding](https://arxiv.org/abs/2210.03347) by Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova.
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
 1. **[PoolFormer](https://huggingface.co/docs/transformers/model_doc/poolformer)** (from Sea AI Labs) released with the paper [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418) by Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng.
-1. **[Pop2Piano](https://huggingface.co/docs/transformers/main/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
+1. **[Pop2Piano](https://huggingface.co/docs/transformers/model_doc/pop2piano)** released with the paper [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi, Kyogu Lee. 
 1. **[ProphetNet](https://huggingface.co/docs/transformers/model_doc/prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
 1. **[PVT](https://huggingface.co/docs/transformers/model_doc/pvt)** (from Nanjing University, The University of Hong Kong etc.) released with the paper [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://arxiv.org/pdf/2102.12122.pdf) by Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao.
 1. **[QDQBert](https://huggingface.co/docs/transformers/model_doc/qdqbert)** (from NVIDIA) released with the paper [Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation](https://arxiv.org/abs/2004.09602) by Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius.
@@ -448,9 +451,10 @@ conda install -c huggingface transformers
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[ViT Hybrid](https://huggingface.co/docs/transformers/model_doc/vit_hybrid)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
-1. **[VitDet](https://huggingface.co/docs/transformers/main/model_doc/vitdet)** (from Meta AI) released with the paper [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) by Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.
+1. **[VitDet](https://huggingface.co/docs/transformers/model_doc/vitdet)** (from Meta AI) released with the paper [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) by Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
 1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (from Meta AI) released with the paper [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas.
+1. **[VITS](https://huggingface.co/docs/transformers/model_doc/vits)** (from Kakao Enterprise) released with the paper [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103) by Jaehyeon Kim, Jungil Kong, Juhee Son.
 1. **[ViViT](https://huggingface.co/docs/transformers/model_doc/vivit)** (from Google Research) released with the paper [ViViT: A Video Vision Transformer](https://arxiv.org/abs/2103.15691) by Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (from Facebook AI) released 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, Sravya Popuri, Dmytro Okhonko, Juan Pino.
@@ -469,7 +473,7 @@ conda install -c huggingface transformers
 1. **[XLS-R](https://huggingface.co/docs/transformers/model_doc/xls_r)** (from Facebook AI) released with the paper [XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale](https://arxiv.org/abs/2111.09296) by Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, Kritika Singh, Patrick von Platen, Yatharth Saraf, Juan Pino, Alexei Baevski, Alexis Conneau, Michael Auli.
 1. **[XLSR-Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/xlsr_wav2vec2)** (from Facebook AI) released with the paper [Unsupervised Cross-Lingual Representation Learning For Speech Recognition](https://arxiv.org/abs/2006.13979) by Alexis Conneau, Alexei Baevski, Ronan Collobert, Abdelrahman Mohamed, Michael Auli.
 1. **[YOLOS](https://huggingface.co/docs/transformers/model_doc/yolos)** (from Huazhong University of Science & Technology) released with the paper [You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection](https://arxiv.org/abs/2106.00666) by Yuxin Fang, Bencheng Liao, Xinggang Wang, Jiemin Fang, Jiyang Qi, Rui Wu, Jianwei Niu, Wenyu Liu.
-1. **[YOSO](https://huggingface.co/docs/transformers/model_doc/yoso)** (from the University of Wisconsin - Madison) released with the paper [You Only Sample (Almost) by Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh.
+1. **[YOSO](https://huggingface.co/docs/transformers/model_doc/yoso)** (from the University of Wisconsin - Madison) released with the paper [You Only Sample (Almost) Once: Linear Cost Self-Attention Via Bernoulli Sampling](https://arxiv.org/abs/2111.09714) by Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh.
 1. 想要貢獻新的模型？我們這裡有一份**詳細指引和模板**來引導你加入新的模型。你可以在 [`templates`](./templates) 目錄中找到它們。記得查看[貢獻指引](./CONTRIBUTING.md)並在開始寫 PR 前聯繫維護人員或開一個新的 issue 來獲得 feedbacks。
 
 要檢查某個模型是否已有 Flax、PyTorch 或 TensorFlow 的實作，或其是否在🤗 Tokenizers 函式庫中有對應的 tokenizer，敬請參閱[此表](https://huggingface.co/docs/transformers/index#supported-frameworks)。

docs/README.md

@@ -81,7 +81,7 @@ The `preview` command only works with existing doc files. When you add a complet
 
 ## Adding a new element to the navigation bar
 
-Accepted files are Markdown (.md or .md).
+Accepted files are Markdown (.md).
 
 Create a file with its extension and put it in the source directory. You can then link it to the toc-tree by putting
 the filename without the extension in the [`_toctree.yml`](https://github.com/huggingface/transformers/blob/main/docs/source/en/_toctree.yml) file.

docs/source/en/_toctree.yml

@@ -98,6 +98,8 @@
     title: Use model-specific APIs
   - local: custom_models
     title: Share a custom model
+  - local: chat_templating
+    title: Templates for chat models
   - local: sagemaker
     title: Run training on Amazon SageMaker
   - local: serialization
@@ -313,6 +315,8 @@
         title: ErnieM
       - local: model_doc/esm
         title: ESM
+      - local: model_doc/falcon
+        title: Falcon
       - local: model_doc/flan-t5
         title: FLAN-T5
       - local: model_doc/flan-ul2
@@ -405,6 +409,8 @@
         title: Pegasus
       - local: model_doc/pegasus_x
         title: PEGASUS-X
+      - local: model_doc/persimmon
+        title: Persimmon
       - local: model_doc/phobert
         title: PhoBERT
       - local: model_doc/plbart
@@ -603,6 +609,8 @@
         title: UniSpeech
       - local: model_doc/unispeech-sat
         title: UniSpeech-SAT
+      - local: model_doc/vits
+        title: VITS
       - local: model_doc/wav2vec2
         title: Wav2Vec2
       - local: model_doc/wav2vec2-conformer
@@ -630,6 +638,8 @@
         title: BLIP-2
       - local: model_doc/bridgetower
         title: BridgeTower
+      - local: model_doc/bros
+        title: BROS
       - local: model_doc/chinese_clip
         title: Chinese-CLIP
       - local: model_doc/clip

docs/source/en/accelerate.md

@@ -133,4 +133,4 @@ accelerate launch train.py
 >>> notebook_launcher(training_function)
 ```
 
-For more information about 🤗 Accelerate and it's rich features, refer to the [documentation](https://huggingface.co/docs/accelerate).
+For more information about 🤗 Accelerate and its rich features, refer to the [documentation](https://huggingface.co/docs/accelerate).

docs/source/en/add_new_model.md

@@ -361,7 +361,7 @@ We expect that every model added to 🤗 Transformers passes a couple of integra
 model and the reimplemented version in 🤗 Transformers have to give the exact same output up to a precision of 0.001!
 Since it is normal that the exact same model written in different libraries can give a slightly different output
 depending on the library framework, we accept an error tolerance of 1e-3 (0.001). It is not enough if the model gives
-nearly the same output, they have to be the almost identical. Therefore, you will certainly compare the intermediate
+nearly the same output, they have to be almost identical. Therefore, you will certainly compare the intermediate
 outputs of the 🤗 Transformers version multiple times against the intermediate outputs of the original implementation of
 *brand_new_bert* in which case an **efficient** debugging environment of the original repository is absolutely
 important. Here is some advice is to make your debugging environment as efficient as possible.

docs/source/en/add_tensorflow_model.md

@@ -56,7 +56,7 @@ you might recall from our [general overview of 🤗 Transformers](add_new_model#
 that we are an opinionated bunch - the ease of use of 🤗 Transformers relies on consistent design choices. From
 experience, we can tell you a few important things about adding TensorFlow models:
 
-- Don't reinvent the wheel! More often that not, there are at least two reference implementations you should check: the
+- Don't reinvent the wheel! More often than not, there are at least two reference implementations you should check: the
 PyTorch equivalent of the model you are implementing and other TensorFlow models for the same class of problems.
 - Great model implementations survive the test of time. This doesn't happen because the code is pretty, but rather
 because the code is clear, easy to debug and build upon. If you make the life of the maintainers easy with your
@@ -101,7 +101,7 @@ TensorFlow-related pull request.
 
 **2. Prepare transformers dev environment**
 
-Having selected the model architecture, open an draft PR to signal your intention to work on it. Follow the
+Having selected the model architecture, open a draft PR to signal your intention to work on it. Follow the
 instructions below to set up your environment and open a draft PR.
 
 1. Fork the [repository](https://github.com/huggingface/transformers) by clicking on the 'Fork' button on the
@@ -328,7 +328,7 @@ That's it! 🎉
 ## Debugging mismatches across ML frameworks 🐛
 
 At some point, when adding a new architecture or when creating TensorFlow weights for an existing architecture, you
-might come across errors compaining about mismatches between PyTorch and TensorFlow. You might even decide to open the
+might come across errors complaining about mismatches between PyTorch and TensorFlow. You might even decide to open the
 model architecture code for the two frameworks, and find that they look identical. What's going on? 🤔
 
 First of all, let's talk about why understanding these mismatches matters. Many community members will use 🤗
@@ -351,7 +351,7 @@ ingredient here is patience. Here is our suggested workflow for when you come ac
    that you'll have to venture into the source implementation of said instruction. In some cases, you might find an
    issue with a reference implementation - don't abstain from opening an issue in the upstream repository.
 
-In some cases, in dicussion with the 🤗 Transformers team, we might find that the fixing the mismatch is infeasible.
+In some cases, in discussion with the 🤗 Transformers team, we might find that fixing the mismatch is infeasible.
 When the mismatch is very small in the output layers of the model (but potentially large in the hidden states), we
 might decide to ignore it in favor of distributing the model. The `pt-to-tf` CLI mentioned above has a `--max-error`
 flag to override the error message at weight conversion time.

docs/source/en/autoclass_tutorial.md

@@ -16,7 +16,7 @@ rendered properly in your Markdown viewer.
 
 # Load pretrained instances with an AutoClass
 
-With so many different Transformer architectures, it can be challenging to create one for your checkpoint. As a part of 🤗 Transformers core philosophy to make the library easy, simple and flexible to use, an `AutoClass` automatically infer and load the correct architecture from a given checkpoint. The `from_pretrained()` method lets you quickly load a pretrained model for any architecture so you don't have to devote time and resources to train a model from scratch. Producing this type of checkpoint-agnostic code means if your code works for one checkpoint, it will work with another checkpoint - as long as it was trained for a similar task - even if the architecture is different.
+With so many different Transformer architectures, it can be challenging to create one for your checkpoint. As a part of 🤗 Transformers core philosophy to make the library easy, simple and flexible to use, an `AutoClass` automatically infers and loads the correct architecture from a given checkpoint. The `from_pretrained()` method lets you quickly load a pretrained model for any architecture so you don't have to devote time and resources to train a model from scratch. Producing this type of checkpoint-agnostic code means if your code works for one checkpoint, it will work with another checkpoint - as long as it was trained for a similar task - even if the architecture is different.
 
 <Tip>
 

docs/source/en/big_models.md

@@ -23,11 +23,11 @@ from PyTorch is:
 2. Load your pretrained weights.
 3. Put those pretrained weights in your random model.
 
-Step 1 and 2 both require a full version of the model in memory, which is not a problem in most cases, but if your model starts weighing several GigaBytes, those two copies can make you got our of RAM. Even worse, if you are using `torch.distributed` to launch a distributed training, each process will load the pretrained model and store these two copies in RAM.
+Step 1 and 2 both require a full version of the model in memory, which is not a problem in most cases, but if your model starts weighing several GigaBytes, those two copies can make you get out of RAM. Even worse, if you are using `torch.distributed` to launch a distributed training, each process will load the pretrained model and store these two copies in RAM.
 
 <Tip>
 
-Note that the randomly created model is initialized with "empty" tensors, which take the space in memory without filling it (thus the random values are whatever was in this chunk of memory at a given time). The random initialization following the appropriate distribution for the kind of model/parameters instatiated (like a normal distribution for instance) is only performed after step 3 on the non-initialized weights, to be as fast as possible! 
+Note that the randomly created model is initialized with "empty" tensors, which take the space in memory without filling it (thus the random values are whatever was in this chunk of memory at a given time). The random initialization following the appropriate distribution for the kind of model/parameters instantiated (like a normal distribution for instance) is only performed after step 3 on the non-initialized weights, to be as fast as possible! 
 
 </Tip>
 
@@ -120,4 +120,4 @@ If you want to directly load such a sharded checkpoint inside a model without us
 
 Sharded checkpoints reduce the memory usage during step 2 of the workflow mentioned above, but in order to use that model in a low memory setting, we recommend leveraging our tools based on the Accelerate library.
 
-Please read the following guide for more information: [Large model loading using Accelerate](./main_classes/model#large-model-loading)
+Please read the following guide for more information: [Large model loading using Accelerate](./main_classes/model#large-model-loading)

docs/source/en/chat_templating.md

@@ -0,0 +1,255 @@
+<!--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.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+# Templates for Chat Models
+
+## Introduction
+
+An increasingly common use case for LLMs is **chat**. In a chat context, rather than continuing a single string
+of text (as is the case with a standard language model), the model instead continues a conversation that consists
+of one or more **messages**, each of which includes a **role** as well as message text.
+
+Most commonly, these roles are "user" for messages sent by the user, and "assistant" for messages sent by the model.
+Some models also support a "system" role. System messages are usually sent at the beginning of the conversation
+and include directives about how the model should behave in the subsequent chat.
+
+All language models, including models fine-tuned for chat, operate on linear sequences of tokens and do not intrinsically
+have special handling for roles. This means that role information is usually injected by adding control tokens
+between messages, to indicate both the message boundary and the relevant roles.
+
+Unfortunately, there isn't (yet!) a standard for which tokens to use, and so different models have been trained
+with wildly different formatting and control tokens for chat. This can be a real problem for users - if you use the
+wrong format, then the model will be confused by your input, and your performance will be a lot worse than it should be.
+This is the problem that **chat templates** aim to resolve. 
+
+Chat conversations are typically represented as a list of dictionaries, where each dictionary contains `role`
+and `content` keys, and represents a single chat message. Chat templates are strings containing a Jinja template that
+specifies how to format a conversation for a given model into a single tokenizable sequence. By storing this information
+with the tokenizer, we can ensure that models get input data in the format they expect.
+
+Let's make this concrete with a quick example using the `BlenderBot` model. BlenderBot has an extremely simple default 
+template, which mostly just adds whitespace between rounds of dialogue:
+
+```python
+>>> from transformers import AutoTokenizer
+>>> tokenizer = AutoTokenizer.from_pretrained("facebook/blenderbot-400M-distill")
+
+>>> chat = [
+...   {"role": "user", "content": "Hello, how are you?"},
+...   {"role": "assistant", "content": "I'm doing great. How can I help you today?"},
+...   {"role": "user", "content": "I'd like to show off how chat templating works!"},
+... ]
+
+>>> tokenizer.apply_chat_template(chat, tokenize=False)
+" Hello, how are you?  I'm doing great. How can I help you today?   I'd like to show off how chat templating works!</s>"
+```
+
+Notice how the entire chat is condensed into a single string. If we use `tokenize=True`, which is the default setting,
+that string will also be tokenized for us. To see a more complex template in action, though, let's use the 
+`meta-llama/Llama-2-7b-chat-hf` model. Note that this model has gated access, so you will have to
+[request access on the repo](https://huggingface.co/meta-llama/Llama-2-7b-chat-hf) if you want to run this code yourself:
+
+```python
+>> from transformers import AutoTokenizer
+>> tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-chat-hf")
+
+>> chat = [
+...   {"role": "user", "content": "Hello, how are you?"},
+...   {"role": "assistant", "content": "I'm doing great. How can I help you today?"},
+...   {"role": "user", "content": "I'd like to show off how chat templating works!"},
+... ]
+
+>> tokenizer.use_default_system_prompt = False
+>> tokenizer.apply_chat_template(chat, tokenize=False)
+"<s>[INST] Hello, how are you? [/INST] I'm doing great. How can I help you today? </s><s>[INST] I'd like to show off how chat templating works! [/INST]"
+```
+
+Note that this time, the tokenizer has added the control tokens [INST] and [/INST] to indicate the start and end of 
+user messages (but not assistant messages!)
+
+## How do chat templates work?
+
+The chat template for a model is stored on the `tokenizer.chat_template` attribute. If no chat template is set, the
+default template for that model class is used instead. Let's take a look at the template for `BlenderBot`:
+
+```python
+
+>>> from transformers import AutoTokenizer
+>>> tokenizer = AutoTokenizer.from_pretrained("facebook/blenderbot-400M-distill")
+
+>>> tokenizer.default_chat_template
+"{% for message in messages %}{% if message['role'] == 'user' %}{{ ' ' }}{% endif %}{{ message['content'] }}{% if not loop.last %}{{ '  ' }}{% endif %}{% endfor %}{{ eos_token }}"
+```
+
+That's kind of intimidating. Let's add some newlines and indentation to make it more readable. Note that
+we remove the first newline after each block as well as any preceding whitespace before a block by default, using the 
+Jinja `trim_blocks` and `lstrip_blocks` flags. This means that you can write your templates with indentations and 
+newlines and still have them function correctly!
+
+```
+{% for message in messages %}
+    {% if message['role'] == 'user' %}
+        {{ ' ' }}
+    {% endif %}
+    {{ message['content'] }}
+    {% if not loop.last %}
+        {{ '  ' }}
+    {% endif %}
+{% endfor %}
+{{ eos_token }}
+```
+
+If you've never seen one of these before, this is a [Jinja template](https://jinja.palletsprojects.com/en/3.1.x/templates/).
+Jinja is a templating language that allows you to write simple code that generates text. In many ways, the code and
+syntax resembles Python. In pure Python, this template would look something like this:
+
+```python
+for idx, message in enumerate(messages):
+    if message['role'] == 'user':
+        print(' ')
+    print(message['content'])
+    if not idx == len(messages) - 1:  # Check for the last message in the conversation
+        print('  ')
+print(eos_token)
+```
+
+Effectively, the template does three things:
+1. For each message, if the message is a user message, add a blank space before it, otherwise print nothing.
+2. Add the message content
+3. If the message is not the last message, add two spaces after it. After the final message, print the EOS token.
+
+This is a pretty simple template - it doesn't add any control tokens, and it doesn't support "system" messages, which 
+are a common way to give the model directives about how it should behave in the subsequent conversation.
+But Jinja gives you a lot of flexibility to do those things! Let's see a Jinja template that can format inputs
+similarly to the way LLaMA formats them (note that the real LLaMA template includes handling for default system
+messages and slightly different system message handling in general - don't use this one in your actual code!)
+
+```
+{% for message in messages %}
+    {% if message['role'] == 'user' %}
+        {{ bos_token + '[INST] ' + message['content'] + ' [/INST]' }}
+    {% elif message['role'] == 'system' %}
+        {{ '<<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}
+    {% elif message['role'] == 'assistant' %}
+        {{ ' '  + message['content'] + ' ' + eos_token }}
+    {% endif %}
+{% endfor %}
+```
+
+Hopefully if you stare at this for a little bit you can see what this template is doing - it adds specific tokens based
+on the "role" of each message, which represents who sent it. User, assistant and system messages are clearly
+distinguishable to the model because of the tokens they're wrapped in.
+
+## How do I create a chat template?
+
+Simple, just write a jinja template and set `tokenizer.chat_template`. You may find it easier to start with an 
+existing template from another model and simply edit it for your needs! For example, we could take the LLaMA template
+above and add "[ASST]" and "[/ASST]" to assistant messages:
+
+```
+{% for message in messages %}
+    {% if message['role'] == 'user' %}
+        {{ bos_token + '[INST] ' + message['content'].strip() + ' [/INST]' }}
+    {% elif message['role'] == 'system' %}
+        {{ '<<SYS>>\\n' + message['content'].strip() + '\\n<</SYS>>\\n\\n' }}
+    {% elif message['role'] == 'assistant' %}
+        {{ '[ASST] '  + message['content'] + ' [/ASST]' + eos_token }}
+    {% endif %}
+{% endfor %}
+```
+
+Now, simply set the `tokenizer.chat_template` attribute. Next time you use [`~PreTrainedTokenizer.apply_chat_template`], it will
+use your new template! This attribute will be saved in the `tokenizer_config.json` file, so you can use
+[`~utils.PushToHubMixin.push_to_hub`] to upload your new template to the Hub and make sure everyone's using the right
+template for your model!
+
+```python
+template = tokenizer.chat_template
+template = template.replace("SYS", "SYSTEM")  # Change the system token
+tokenizer.chat_template = template  # Set the new template
+tokenizer.push_to_hub("model_name")  # Upload your new template to the Hub!
+```
+
+The method [`~PreTrainedTokenizer.apply_chat_template`] which uses your chat template is called by the [`ConversationalPipeline`] class, so 
+once you set the correct chat template, your model will automatically become compatible with [`ConversationalPipeline`].
+
+## What are "default" templates?
+
+Before the introduction of chat templates, chat handling was hardcoded at the model class level. For backwards 
+compatibility, we have retained this class-specific handling as default templates, also set at the class level. If a
+model does not have a chat template set, but there is a default template for its model class, the `ConversationalPipeline`
+class and methods like `apply_chat_template` will use the class template instead. You can find out what the default
+template for your tokenizer is by checking the `tokenizer.default_chat_template` attribute.
+
+This is something we do purely for backward compatibility reasons, to avoid breaking any existing workflows. Even when
+the class template is appropriate for your model, we strongly recommend overriding the default template by
+setting the `chat_template` attribute explicitly to make it clear to users that your model has been correctly configured
+for chat, and to future-proof in case the default templates are ever altered or deprecated.
+
+## What template should I use?
+
+When setting the template for a model that's already been trained for chat, you should ensure that the template
+exactly matches the message formatting that the model saw during training, or else you will probably experience
+performance degradation. This is true even if you're training the model further - you will probably get the best 
+performance if you keep the chat tokens constant. This is very analogous to tokenization - you generally get the
+best performance for inference or fine-tuning when you precisely match the tokenization used during training.
+
+If you're training a model from scratch, or fine-tuning a base language model for chat, on the other hand,
+you have a lot of freedom to choose an appropriate template! LLMs are smart enough to learn to handle lots of different
+input formats. Our default template for models that don't have a class-specific template follows the 
+[ChatML format](https://github.com/openai/openai-python/blob/main/chatml.md), and this is a good, flexible choice for many use-cases. It looks like this:
+
+```
+{% for message in messages %}
+    {{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}
+{% endfor %}
+```
+
+If you like this one, here it is in one-liner form, ready to copy into your code:
+
+```
+tokenizer.chat_template = "{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}"
+```
+
+This template wraps each message in `<|im_start|>` and `<|im_end|>` tokens, and simply writes the role as a string, which
+allows for flexibility in the roles you train with. The output looks like this:
+
+```
+<|im_start|>system
+You are a helpful chatbot that will do its best not to say anything so stupid that people tweet about it.<|im_end|>
+<|im_start|>user
+How are you?<|im_end|>
+<|im_start|>assistant
+I'm doing great!<|im_end|>
+```
+
+The "user", "system" and "assistant" roles are the standard for chat, and we recommend using them when it makes sense,
+particularly if you want your model to operate well with [`ConversationalPipeline`]. However, you are not limited
+to these roles - templating is extremely flexible, and any string can be a role.
+
+## I want to use chat templates! How should I get started?
+
+If you have any chat models, you should set their `tokenizer.chat_template` attribute and test it using
+[`~PreTrainedTokenizer.apply_chat_template`]. This applies even if you're not the model owner - if you're using a model
+with an empty chat template, or one that's still using the default class template, please open a [pull request](https://huggingface.co/docs/hub/repositories-pull-requests-discussions) to
+the model repository so that this attribute can be set properly!
+
+Once the attribute is set, that's it, you're done! `tokenizer.apply_chat_template` will now work correctly for that
+model, which means it is also automatically supported in places like `ConversationalPipeline`!
+
+By ensuring that models have this attribute, we can make sure that the whole community gets to use the full power of
+open-source models. Formatting mismatches have been haunting the field and silently harming performance for too long - 
+it's time to put an end to them!

docs/source/en/community.md

@@ -1,4 +1,4 @@
-<!--⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+<!--⚠️ Note that this file is in Markdown but contains specific syntax for our doc-builder (similar to MDX) that may not be
 rendered properly in your Markdown viewer.
 -->
 
@@ -10,7 +10,7 @@ This page regroups resources around 🤗 Transformers developed by the community
 
 | Resource     |      Description      |      Author      |
 |:----------|:-------------|------:|
-| [Hugging Face Transformers Glossary Flashcards](https://www.darigovresearch.com/huggingface-transformers-glossary-flashcards) | A set of flashcards based on the [Transformers Docs Glossary](glossary) that has been put into a form which can be easily learnt/revised using [Anki ](https://apps.ankiweb.net/) an open source, cross platform app specifically designed for long term knowledge retention. See this [Introductory video on how to use the flashcards](https://www.youtube.com/watch?v=Dji_h7PILrw). | [Darigov Research](https://www.darigovresearch.com/) |
+| [Hugging Face Transformers Glossary Flashcards](https://www.darigovresearch.com/huggingface-transformers-glossary-flashcards) | A set of flashcards based on the [Transformers Docs Glossary](glossary) that has been put into a form which can be easily learned/revised using [Anki ](https://apps.ankiweb.net/) an open source, cross platform app specifically designed for long term knowledge retention. See this [Introductory video on how to use the flashcards](https://www.youtube.com/watch?v=Dji_h7PILrw). | [Darigov Research](https://www.darigovresearch.com/) |
 
 ## Community notebooks:
 
@@ -35,7 +35,7 @@ This page regroups resources around 🤗 Transformers developed by the community
 |[Speed up Fine-Tuning in Transformers with Dynamic Padding / Bucketing](https://github.com/ELS-RD/transformers-notebook/blob/master/Divide_Hugging_Face_Transformers_training_time_by_2_or_more.ipynb)|How to speed up fine-tuning by a factor of 2 using dynamic padding / bucketing|[Michael Benesty](https://github.com/pommedeterresautee) |[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1CBfRU1zbfu7-ijiOqAAQUA-RJaxfcJoO?usp=sharing)|
 |[Pretrain Reformer for Masked Language Modeling](https://github.com/patrickvonplaten/notebooks/blob/master/Reformer_For_Masked_LM.ipynb)| How to train a Reformer model with bi-directional self-attention layers | [Patrick von Platen](https://github.com/patrickvonplaten) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1tzzh0i8PgDQGV3SMFUGxM7_gGae3K-uW?usp=sharing)|
 |[Expand and Fine Tune Sci-BERT](https://github.com/lordtt13/word-embeddings/blob/master/COVID-19%20Research%20Data/COVID-SciBERT.ipynb)| How to increase vocabulary of a pretrained SciBERT model from AllenAI on the CORD dataset and pipeline it. | [Tanmay Thakur](https://github.com/lordtt13) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1rqAR40goxbAfez1xvF3hBJphSCsvXmh8)|
-|[Fine Tune BlenderBotSmall for Summarization using the Trainer API](https://github.com/lordtt13/transformers-experiments/blob/master/Custom%20Tasks/fine-tune-blenderbot_small-for-summarization.ipynb)| How to fine tune BlenderBotSmall for summarization on a custom dataset, using the Trainer API. | [Tanmay Thakur](https://github.com/lordtt13) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/19Wmupuls7mykSGyRN_Qo6lPQhgp56ymq?usp=sharing)|
+|[Fine Tune BlenderBotSmall for Summarization using the Trainer API](https://github.com/lordtt13/transformers-experiments/blob/master/Custom%20Tasks/fine-tune-blenderbot_small-for-summarization.ipynb)| How to fine-tune BlenderBotSmall for summarization on a custom dataset, using the Trainer API. | [Tanmay Thakur](https://github.com/lordtt13) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/19Wmupuls7mykSGyRN_Qo6lPQhgp56ymq?usp=sharing)|
 |[Fine-tune Electra and interpret with Integrated Gradients](https://github.com/elsanns/xai-nlp-notebooks/blob/master/electra_fine_tune_interpret_captum_ig.ipynb) | How to fine-tune Electra for sentiment analysis and interpret predictions with Captum Integrated Gradients | [Eliza Szczechla](https://elsanns.github.io) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/elsanns/xai-nlp-notebooks/blob/master/electra_fine_tune_interpret_captum_ig.ipynb)|
 |[fine-tune a non-English GPT-2 Model with Trainer class](https://github.com/philschmid/fine-tune-GPT-2/blob/master/Fine_tune_a_non_English_GPT_2_Model_with_Huggingface.ipynb) | How to fine-tune a non-English GPT-2 Model with Trainer class | [Philipp Schmid](https://www.philschmid.de) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/philschmid/fine-tune-GPT-2/blob/master/Fine_tune_a_non_English_GPT_2_Model_with_Huggingface.ipynb)|
 |[Fine-tune a DistilBERT Model for Multi Label Classification task](https://github.com/DhavalTaunk08/Transformers_scripts/blob/master/Transformers_multilabel_distilbert.ipynb) | How to fine-tune a DistilBERT Model for Multi Label Classification task | [Dhaval Taunk](https://github.com/DhavalTaunk08) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/DhavalTaunk08/Transformers_scripts/blob/master/Transformers_multilabel_distilbert.ipynb)|

docs/source/en/create_a_model.md

@@ -209,7 +209,7 @@ Easily reuse this checkpoint for another task by switching to a different model
 The last base class you need before using a model for textual data is a [tokenizer](main_classes/tokenizer) to convert raw text to tensors. There are two types of tokenizers you can use with 🤗 Transformers:
 
 - [`PreTrainedTokenizer`]: a Python implementation of a tokenizer.
-- [`PreTrainedTokenizerFast`]: a tokenizer from our Rust-based [🤗 Tokenizer](https://huggingface.co/docs/tokenizers/python/latest/) library. This tokenizer type is significantly faster - especially during batch tokenization - due to it's Rust implementation. The fast tokenizer also offers additional methods like *offset mapping* which maps tokens to their original words or characters.
+- [`PreTrainedTokenizerFast`]: a tokenizer from our Rust-based [🤗 Tokenizer](https://huggingface.co/docs/tokenizers/python/latest/) library. This tokenizer type is significantly faster - especially during batch tokenization - due to its Rust implementation. The fast tokenizer also offers additional methods like *offset mapping* which maps tokens to their original words or characters.
 
 Both tokenizers support common methods such as encoding and decoding, adding new tokens, and managing special tokens.
 

docs/source/en/custom_models.md

@@ -341,7 +341,7 @@ model. This is different from pushing the code to the Hub in the sense that user
 get the custom models (contrarily to automatically downloading the model code from the Hub).
 
 As long as your config has a `model_type` attribute that is different from existing model types, and that your model
-classes have the right `config_class` attributes, you can just add them to the auto classes likes this:
+classes have the right `config_class` attributes, you can just add them to the auto classes like this:
 
 ```py
 from transformers import AutoConfig, AutoModel, AutoModelForImageClassification

docs/source/en/custom_tools.md

@@ -25,7 +25,7 @@ If you are not aware of what tools and agents are in the context of transformers
 
 <Tip warning={true}>
 
-Transformers Agent is an experimental API that is subject to change at any time. Results returned by the agents
+Transformers Agents is an experimental API that is subject to change at any time. Results returned by the agents
 can vary as the APIs or underlying models are prone to change.
 
 </Tip>

docs/source/en/generation_strategies.md

@@ -55,12 +55,10 @@ When you load a model explicitly, you can inspect the generation configuration t
 >>> from transformers import AutoModelForCausalLM
 
 >>> model = AutoModelForCausalLM.from_pretrained("distilgpt2")
->>> model.generation_config  # doctest: +IGNORE_RESULT
+>>> model.generation_config
 GenerationConfig {
-    "_from_model_config": true,
     "bos_token_id": 50256,
     "eos_token_id": 50256,
-    "transformers_version": "4.26.0.dev0"
 }
 ```
 
@@ -92,7 +90,7 @@ sequences that start with a lower probability initial tokens and would've been i
 - `do_sample`: if set to `True`, this parameter enables decoding strategies such as multinomial sampling, beam-search
 multinomial sampling, Top-K sampling and Top-p sampling. All these strategies select the next token from the probability
 distribution over the entire vocabulary with various strategy-specific adjustments.
-- `num_return_sequences`: the number of sequence candidates to return for each input. This options is only available for
+- `num_return_sequences`: the number of sequence candidates to return for each input. This option is only available for
 the decoding strategies that support multiple sequence candidates, e.g. variations of beam search and sampling. Decoding
 strategies like greedy search and contrastive search return a single output sequence.
 
@@ -146,7 +144,7 @@ one for summarization with beam search). You must have the right Hub permissions
 
 ## Streaming
 
-The `generate()` supports streaming, through its `streamer` input. The `streamer` input is compatible any instance
+The `generate()` supports streaming, through its `streamer` input. The `streamer` input is compatible with any instance
 from a class that has the following methods: `put()` and `end()`. Internally, `put()` is used to push new tokens and
 `end()` is used to flag the end of text generation.
 
@@ -301,7 +299,7 @@ the `num_beams` greater than 1, and set `do_sample=True` to use this decoding st
 The diverse beam search decoding strategy is an extension of the beam search strategy that allows for generating a more diverse
 set of beam sequences to choose from. To learn how it works, refer to [Diverse Beam Search: Decoding Diverse Solutions from Neural Sequence Models](https://arxiv.org/pdf/1610.02424.pdf).
 This approach has three main parameters: `num_beams`, `num_beam_groups`, and `diversity_penalty`.
-The diversily penalty ensures the outputs are distinct across groups, and beam search is used within each group.
+The diversity penalty ensures the outputs are distinct across groups, and beam search is used within each group.
 
 
 ```python
@@ -332,7 +330,8 @@ The diversily penalty ensures the outputs are distinct across groups, and beam s
 
 >>> outputs = model.generate(**inputs, num_beams=5, num_beam_groups=5, max_new_tokens=30, diversity_penalty=1.0)
 >>> tokenizer.decode(outputs[0], skip_special_tokens=True)
-'The aim of this project is to create a new type of living system, one that is more sustainable and efficient than the current one.'
+'The Design Principles are a set of universal design principles that can be applied to any location, climate and
+culture, and they allow us to design the'
 ```
 
 This guide illustrates the main parameters that enable various decoding strategies. More advanced parameters exist for the
@@ -366,7 +365,7 @@ To enable assisted decoding, set the `assistant_model` argument with a model.
 ['Alice and Bob are sitting in a bar. Alice is drinking a beer and Bob is drinking a']
 ```
 
-When using assisted decoding with sampling methods, you can use the `temperarure` argument to control the randomness
+When using assisted decoding with sampling methods, you can use the `temperature` argument to control the randomness
 just like in multinomial sampling. However, in assisted decoding, reducing the temperature will help improving latency.
 
 ```python

docs/source/en/glossary.md

@@ -187,7 +187,7 @@ The model head refers to the last layer of a neural network that accepts the raw
 
 ### image patch
 
-Vision-based Transformers models split an image into smaller patches which are linearly embedded, and then passed as a sequence to the model. You can find the `patch_size` - or resolution - of the model in it's configuration.
+Vision-based Transformers models split an image into smaller patches which are linearly embedded, and then passed as a sequence to the model. You can find the `patch_size` - or resolution - of the model in its configuration.
 
 ### inference
 

docs/source/en/hpo_train.md

@@ -54,6 +54,18 @@ For optuna, see optuna [object_parameter](https://optuna.readthedocs.io/en/stabl
 ...     }
 ```
 
+Optuna provides multi-objective HPO. You can pass `direction` in `hyperparameter_search` and define your own compute_objective to return multiple objective values. The Pareto Front (`List[BestRun]`) will be returned in hyperparameter_search, you should refer to the test case `TrainerHyperParameterMultiObjectOptunaIntegrationTest` in [test_trainer](https://github.com/huggingface/transformers/blob/main/tests/trainer/test_trainer.py). It's like following
+
+```py
+>>> best_trials = trainer.hyperparameter_search(
+...     direction=["minimize", "maximize"],
+...     backend="optuna",
+...     hp_space=optuna_hp_space,
+...     n_trials=20,
+...     compute_objective=compute_objective,
+... )
+```
+
 For raytune, see raytune [object_parameter](https://docs.ray.io/en/latest/tune/api/search_space.html), it's like following:
 
 ```py

docs/source/en/index.md

@@ -76,6 +76,7 @@ The documentation is organized into five sections:
 1. **[BLOOM](model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](model_doc/bort)** (from Alexa) released with the paper [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) by Adrian de Wynter and Daniel J. Perry.
 1. **[BridgeTower](model_doc/bridgetower)** (from Harbin Institute of Technology/Microsoft Research Asia/Intel Labs) released with the paper [BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning](https://arxiv.org/abs/2206.08657) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[BROS](model_doc/bros)** (from NAVER CLOVA) released with the paper [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539) by Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park.
 1. **[ByT5](model_doc/byt5)** (from Google Research) released with the paper [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) by Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel.
 1. **[CamemBERT](model_doc/camembert)** (from Inria/Facebook/Sorbonne) released with the paper [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
 1. **[CANINE](model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
@@ -139,6 +140,7 @@ The documentation is organized into five sections:
 1. **[GPTSAN-japanese](model_doc/gptsan-japanese)** released in the repository [tanreinama/GPTSAN](https://github.com/tanreinama/GPTSAN/blob/main/report/model.md) by Toshiyuki Sakamoto(tanreinama).
 1. **[Graphormer](model_doc/graphormer)** (from Microsoft) released with the paper [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234) by Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu.
 1. **[GroupViT](model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
+1. **[HerBERT](model_doc/herbert)** (from Allegro.pl, AGH University of Science and Technology) released with the paper [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf) by Piotr Rybak, Robert Mroczkowski, Janusz Tracz, Ireneusz Gawlik.
 1. **[Hubert](model_doc/hubert)** (from Facebook) released with the paper [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447) by Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed.
 1. **[I-BERT](model_doc/ibert)** (from Berkeley) released with the paper [I-BERT: Integer-only BERT Quantization](https://arxiv.org/abs/2101.01321) by Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer.
 1. **[IDEFICS](model_doc/idefics)** (from HuggingFace) released with the paper [OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents](https://huggingface.co/papers/2306.16527) by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sanh.
@@ -197,6 +199,7 @@ The documentation is organized into five sections:
 1. **[Pegasus](model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
 1. **[PEGASUS-X](model_doc/pegasus_x)** (from Google) released with the paper [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) by Jason Phang, Yao Zhao, and Peter J. Liu.
 1. **[Perceiver IO](model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
+1. **[Persimmon](model_doc/persimmon)** (from ADEPT) released in a [blog post](https://www.adept.ai/blog/persimmon-8b) by Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani.
 1. **[PhoBERT](model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
 1. **[Pix2Struct](model_doc/pix2struct)** (from Google) released with the paper [Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding](https://arxiv.org/abs/2210.03347) by Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova.
 1. **[PLBart](model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
@@ -255,6 +258,7 @@ The documentation is organized into five sections:
 1. **[VitDet](model_doc/vitdet)** (from Meta AI) released with the paper [Exploring Plain Vision Transformer Backbones for Object Detection](https://arxiv.org/abs/2203.16527) by Yanghao Li, Hanzi Mao, Ross Girshick, Kaiming He.
 1. **[ViTMAE](model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
 1. **[ViTMSN](model_doc/vit_msn)** (from Meta AI) released with the paper [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas.
+1. **[VITS](model_doc/vits)** (from Kakao Enterprise) released with the paper [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103) by Jaehyeon Kim, Jungil Kong, Juhee Son.
 1. **[ViViT](model_doc/vivit)** (from Google Research) released with the paper [ViViT: A Video Vision Transformer](https://arxiv.org/abs/2103.15691) by Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid.
 1. **[Wav2Vec2](model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2-Conformer](model_doc/wav2vec2-conformer)** (from Facebook AI) released 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, Sravya Popuri, Dmytro Okhonko, Juan Pino.
@@ -306,6 +310,7 @@ Flax), PyTorch, and/or TensorFlow.
 |            BLIP-2             |       ✅        |         ❌         |      ❌      |
 |             BLOOM             |       ✅        |         ❌         |      ✅      |
 |          BridgeTower          |       ✅        |         ❌         |      ❌      |
+|             BROS              |       ✅        |         ❌         |      ❌      |
 |           CamemBERT           |       ✅        |         ✅         |      ❌      |
 |            CANINE             |       ✅        |         ❌         |      ❌      |
 |         Chinese-CLIP          |       ✅        |         ❌         |      ❌      |
@@ -416,6 +421,7 @@ Flax), PyTorch, and/or TensorFlow.
 |            Pegasus            |       ✅        |         ✅         |      ✅      |
 |           PEGASUS-X           |       ✅        |         ❌         |      ❌      |
 |           Perceiver           |       ✅        |         ❌         |      ❌      |
+|           Persimmon           |       ✅        |         ❌         |      ❌      |
 |          Pix2Struct           |       ✅        |         ❌         |      ❌      |
 |            PLBart             |       ✅        |         ❌         |      ❌      |
 |          PoolFormer           |       ✅        |         ❌         |      ❌      |
@@ -475,6 +481,7 @@ Flax), PyTorch, and/or TensorFlow.
 |            VitDet             |       ✅        |         ❌         |      ❌      |
 |            ViTMAE             |       ✅        |         ✅         |      ❌      |
 |            ViTMSN             |       ✅        |         ❌         |      ❌      |
+|             VITS              |       ✅        |         ❌         |      ❌      |
 |             ViViT             |       ✅        |         ❌         |      ❌      |
 |           Wav2Vec2            |       ✅        |         ✅         |      ✅      |
 |      Wav2Vec2-Conformer       |       ✅        |         ❌         |      ❌      |

docs/source/en/llm_tutorial.md

@@ -27,7 +27,7 @@ This tutorial will show you how to:
 
 * Generate text with an LLM
 * Avoid common pitfalls
-* Next steps to help you get the most out your LLM
+* Next steps to help you get the most out of your LLM
 
 Before you begin, make sure you have all the necessary libraries installed:
 

docs/source/en/main_classes/agent.md

@@ -18,7 +18,7 @@ rendered properly in your Markdown viewer.
 
 <Tip warning={true}>
 
-Transformers Agent is an experimental API which is subject to change at any time. Results returned by the agents
+Transformers Agents is an experimental API which is subject to change at any time. Results returned by the agents
 can vary as the APIs or underlying models are prone to change.
 
 </Tip>

docs/source/en/main_classes/output.md

@@ -40,6 +40,13 @@ an optional `attentions` attribute. Here we have the `loss` since we passed alon
 `hidden_states` and `attentions` because we didn't pass `output_hidden_states=True` or
 `output_attentions=True`.
 
+<Tip>
+
+When passing `output_hidden_states=True` you may expect the `outputs.hidden_states[-1]` to match `outputs.last_hidden_states` exactly.
+However, this is not always the case. Some models apply normalization or subsequent process to the last hidden state when it's returned.
+</Tip>
+
+
 You can access each attribute as you would usually do, and if that attribute has not been returned by the model, you
 will get `None`. Here for instance `outputs.loss` is the loss computed by the model, and `outputs.attentions` is
 `None`.

docs/source/en/main_classes/tokenizer.md

@@ -55,6 +55,7 @@ to a given token).
 
 [[autodoc]] PreTrainedTokenizer
     - __call__
+    - apply_chat_template
     - batch_decode
     - decode
     - encode
@@ -68,6 +69,7 @@ loaded very simply into 🤗 transformers. Take a look at the [Using tokenizers
 
 [[autodoc]] PreTrainedTokenizerFast
     - __call__
+    - apply_chat_template
     - batch_decode
     - decode
     - encode

docs/source/en/model_doc/bros.md

@@ -0,0 +1,115 @@
+<!--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.
+-->
+
+# BROS
+
+## Overview
+
+The BROS model was proposed in [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539) by Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park.
+
+BROS stands for *BERT Relying On Spatiality*. It is an encoder-only Transformer model that takes a sequence of tokens and their bounding boxes as inputs and outputs a sequence of hidden states. BROS encode relative spatial information instead of using absolute spatial information.
+
+It is pre-trained with two objectives: a token-masked language modeling objective (TMLM) used in BERT, and a novel area-masked language modeling objective (AMLM)
+In TMLM, tokens are randomly masked, and the model predicts the masked tokens using spatial information and other unmasked tokens.
+AMLM is a 2D version of TMLM. It randomly masks text tokens and predicts with the same information as TMLM, but it masks text blocks (areas).
+
+`BrosForTokenClassification` has a simple linear layer on top of BrosModel. It predicts the label of each token.
+`BrosSpadeEEForTokenClassification` has an `initial_token_classifier` and `subsequent_token_classifier` on top of BrosModel. `initial_token_classifier` is used to predict the first token of each entity, and `subsequent_token_classifier` is used to predict the next token of within entity. `BrosSpadeELForTokenClassification` has an `entity_linker` on top of BrosModel. `entity_linker` is used to predict the relation between two entities.
+
+`BrosForTokenClassification` and `BrosSpadeEEForTokenClassification` essentially perform the same job. However, `BrosForTokenClassification` assumes input tokens are perfectly serialized (which is very challenging task since they exist in a 2D space), while `BrosSpadeEEForTokenClassification` allows for more flexibility in handling serialization errors as it predicts next connection tokens from one token.
+
+`BrosSpadeELForTokenClassification` perform the intra-entity linking task. It predicts relation from one token (of one entity) to another token (of another entity) if these two entities share some relation.
+
+BROS achieves comparable or better result on Key Information Extraction (KIE) benchmarks such as FUNSD, SROIE, CORD and SciTSR, without relying on explicit visual features.
+
+
+The abstract from the paper is the following:
+
+*Key information extraction (KIE) from document images requires understanding the contextual and spatial semantics of texts in two-dimensional (2D) space. Many recent studies try to solve the task by developing pre-trained language models focusing on combining visual features from document images with texts and their layout. On the other hand, this paper tackles the problem by going back to the basic: effective combination of text and layout. Specifically, we propose a pre-trained language model, named BROS (BERT Relying On Spatiality), that encodes relative positions of texts in 2D space and learns from unlabeled documents with area-masking strategy. With this optimized training scheme for understanding texts in 2D space, BROS shows comparable or better performance compared to previous methods on four KIE benchmarks (FUNSD, SROIE*, CORD, and SciTSR) without relying on visual features. This paper also reveals two real-world challenges in KIE tasks-(1) minimizing the error from incorrect text ordering and (2) efficient learning from fewer downstream examples-and demonstrates the superiority of BROS over previous methods.*
+
+Tips:
+
+- [`~transformers.BrosModel.forward`] requires `input_ids` and `bbox` (bounding box). Each bounding box should be in (x0, y0, x1, y1) format (top-left corner, bottom-right corner). Obtaining of Bounding boxes depends on external OCR system. The `x` coordinate should be normalized by document image width, and the `y` coordinate should be normalized by document image height.
+
+```python
+def expand_and_normalize_bbox(bboxes, doc_width, doc_height):
+    # here, bboxes are numpy array
+
+    # Normalize bbox -> 0 ~ 1
+    bboxes[:, [0, 2]] = bboxes[:, [0, 2]] / width
+    bboxes[:, [1, 3]] = bboxes[:, [1, 3]] / height
+```
+
+- [`~transformers.BrosForTokenClassification.forward`, `~transformers.BrosSpadeEEForTokenClassification.forward`, `~transformers.BrosSpadeEEForTokenClassification.forward`] require not only `input_ids` and `bbox` but also `box_first_token_mask` for loss calculation. It is a mask to filter out non-first tokens of each box. You can obtain this mask by saving start token indices of bounding boxes when creating `input_ids` from words. You can make `box_first_token_mask` with following code,
+
+
+```python
+def make_box_first_token_mask(bboxes, words, tokenizer, max_seq_length=512):
+
+    box_first_token_mask = np.zeros(max_seq_length, dtype=np.bool_)
+
+    # encode(tokenize) each word from words (List[str])
+    input_ids_list: List[List[int]] = [tokenizer.encode(e, add_special_tokens=False) for e in words]
+
+    # get the length of each box
+    tokens_length_list: List[int] = [len(l) for l in input_ids_list]
+
+    box_end_token_indices = np.array(list(itertools.accumulate(tokens_length_list)))
+    box_start_token_indices = box_end_token_indices - np.array(tokens_length_list)
+
+    # filter out the indices that are out of max_seq_length
+    box_end_token_indices = box_end_token_indices[box_end_token_indices < max_seq_length - 1]
+    if len(box_start_token_indices) > len(box_end_token_indices):
+        box_start_token_indices = box_start_token_indices[: len(box_end_token_indices)]
+
+    # set box_start_token_indices to True
+    box_first_token_mask[box_start_token_indices] = True
+
+    return box_first_token_mask
+
+```
+
+- Demo scripts can be found [here](https://github.com/clovaai/bros).
+
+This model was contributed by [jinho8345](https://huggingface.co/jinho8345). The original code can be found [here](https://github.com/clovaai/bros).
+
+## BrosConfig
+
+[[autodoc]] BrosConfig
+
+## BrosProcessor
+
+[[autodoc]] BrosProcessor
+    - __call__
+
+## BrosModel
+
+[[autodoc]] BrosModel
+    - forward
+
+
+## BrosForTokenClassification
+
+[[autodoc]] BrosForTokenClassification
+    - forward
+
+
+## BrosSpadeEEForTokenClassification
+
+[[autodoc]] BrosSpadeEEForTokenClassification
+    - forward
+
+
+## BrosSpadeELForTokenClassification
+
+[[autodoc]] BrosSpadeELForTokenClassification
+    - forward

docs/source/en/model_doc/code_llama.md

@@ -49,6 +49,8 @@ Here is a sample usage
 python src/transformers/models/llama/convert_llama_weights_to_hf.py \
     --input_dir /path/to/downloaded/llama/weights --model_size 7B --output_dir /output/path
 ```
+Note that executing the script requires enough CPU RAM to host the whole model in float16 precision (even if the biggest versions
+come in several checkpoints they each contain a part of each weight of the model, so we need to load them all in RAM).
 
 - After conversion, the model and tokenizer can be loaded via:
 
@@ -90,8 +92,8 @@ If you only want the infilled part:
 >>> generator = pipeline("text-generation",model="codellama/CodeLlama-7b-hf",torch_dtype=torch.float16, device_map="auto")
 >>> generator('def remove_non_ascii(s: str) -> str:\n    """ <FILL_ME>\n    return result', max_new_tokens = 128, return_type = 1)
 ```
-Note that executing the script requires enough CPU RAM to host the whole model in float16 precision (even if the biggest versions
-come in several checkpoints they each contain a part of each weight of the model, so we need to load them all in RAM). For the 75B model, it's thus 145GB of RAM needed.
+
+Under the hood, the tokenizer [automatically splits by `<FILL_ME>`](https://huggingface.co/docs/transformers/main/model_doc/code_llama#transformers.CodeLlamaTokenizer.fill_token) to create a formatted input string that follows [the original training pattern](https://github.com/facebookresearch/codellama/blob/cb51c14ec761370ba2e2bc351374a79265d0465e/llama/generation.py#L402). This is more robust than preparing the pattern yourself: it avoids pitfalls, such as token glueing, that are very hard to debug.  To see how much CPU and GPU memory you need for this model or others, try [this calculator](https://huggingface.co/spaces/hf-accelerate/model-memory-usage) which can help determine that value.
 
 - The LLaMA tokenizer is a BPE model based on [sentencepiece](https://github.com/google/sentencepiece). One quirk of sentencepiece is that when decoding a sequence, if the first token is the start of the word (e.g. "Banana"), the tokenizer does not prepend the prefix space to the string.
 

docs/source/en/model_doc/deberta-v2.md

@@ -152,3 +152,8 @@ contributed by [kamalkraj](https://huggingface.co/kamalkraj). The original code
 
 [[autodoc]] TFDebertaV2ForQuestionAnswering
     - call
+
+## TFDebertaV2ForMultipleChoice
+
+[[autodoc]] TFDebertaV2ForMultipleChoice
+    - call

docs/source/en/model_doc/falcon.md

@@ -0,0 +1,84 @@
+<!--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.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+# Falcon
+
+## Overview
+
+Falcon is a class of causal decoder-only models built by [TII](https://www.tii.ae/). The largest Falcon checkpoints
+have been trained on >=1T tokens of text, with a particular emphasis on the [RefinedWeb](https://arxiv.org/abs/2306.01116)
+corpus. They are made available under the Apache 2.0 license.
+
+
+Falcon's architecture is modern and optimized for inference, with multi-query attention and support for efficient
+attention variants like `FlashAttention`. Both 'base' models trained only as causal language models as well as
+'instruct' models that have received further fine-tuning are available.
+
+
+Falcon models are (as of 2023) some of the largest and most powerful open-source language models,
+and consistently rank highly in the [OpenLLM leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).
+
+## Converting custom checkpoints 
+
+<Tip>
+
+Falcon models were initially added to the Hugging Face Hub as custom code checkpoints. However, Falcon is now fully
+supported in the Transformers library. If you fine-tuned a model from a custom code checkpoint, we recommend converting
+your checkpoint to the new in-library format, as this should give significant improvements to stability and
+performance, especially for generation, as well as removing the need to use `trust_remote_code=True`!
+
+</Tip>
+
+You can convert custom code checkpoints to full Transformers checkpoints using the `convert_custom_code_checkpoint.py` 
+script located in the
+[Falcon model directory](https://github.com/huggingface/transformers/tree/main/src/transformers/models/falcon)
+of the Transformers library. To use this script, simply call it with 
+`python convert_custom_code_checkpoint.py --checkpoint_dir my_model`. This will convert your checkpoint in-place, and
+you can immediately load it from the directory afterwards with e.g. `from_pretrained()`. If your model hasn't been
+uploaded to the Hub, we recommend making a backup before attempting the conversion, just in case!
+
+
+## FalconConfig
+
+[[autodoc]] FalconConfig
+    - all
+
+## FalconModel
+
+[[autodoc]] FalconModel
+    - forward
+
+## FalconForCausalLM
+
+[[autodoc]] FalconForCausalLM
+    - forward
+
+## FalconForSequenceClassification
+
+[[autodoc]] FalconForSequenceClassification
+    - forward
+
+## FalconForTokenClassification
+
+[[autodoc]] FalconForTokenClassification
+    - forward
+
+## FalconForQuestionAnswering
+
+[[autodoc]] FalconForQuestionAnswering
+    - forward
+
+

docs/source/en/model_doc/llama.md

@@ -55,6 +55,28 @@ Based on the original LLaMA model, Meta AI has released some follow-up works:
 
 - **Llama2**: Llama2 is an improved version of Llama with some architectural tweaks (Grouped Query Attention), and is pre-trained on 2Trillion tokens. Refer to the documentation of Llama2 which can be found [here](llama2).
 
+## Resources
+
+A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with LLaMA. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
+
+<PipelineTag pipeline="text-classification"/>
+
+- A [notebook](https://colab.research.google.com/github/bigscience-workshop/petals/blob/main/examples/prompt-tuning-sst2.ipynb#scrollTo=f04ba4d2) on how to use prompt tuning to adapt the LLaMA model for text classification task. 🌎
+
+<PipelineTag pipeline="question-answering"/>
+
+- [StackLLaMA: A hands-on guide to train LLaMA with RLHF](https://huggingface.co/blog/stackllama#stackllama-a-hands-on-guide-to-train-llama-with-rlhf), a blog post about how to train LLaMA to answer questions on [Stack Exchange](https://stackexchange.com/) with RLHF.
+
+⚗️ Optimization
+- A [notebook](https://colab.research.google.com/drive/1SQUXq1AMZPSLD4mk3A3swUIc6Y2dclme?usp=sharing) on how to fine-tune LLaMA model using xturing library on GPU which has limited memory. 🌎 
+
+⚡️ Inference
+- A [notebook](https://colab.research.google.com/github/DominguesM/alpaca-lora-ptbr-7b/blob/main/notebooks/02%20-%20Evaluate.ipynb) on how to run the LLaMA Model using PeftModel from the 🤗 PEFT library. 🌎 
+- A [notebook](https://colab.research.google.com/drive/1l2GiSSPbajVyp2Nk3CFT4t3uH6-5TiBe?usp=sharing) on how to load a PEFT adapter LLaMA model with LangChain. 🌎
+
+🚀 Deploy
+- A [notebook](https://colab.research.google.com/github/lxe/simple-llama-finetuner/blob/master/Simple_LLaMA_FineTuner.ipynb#scrollTo=3PM_DilAZD8T) on how to fine-tune LLaMA model using LoRA method via the 🤗 PEFT library with intuitive UI. 🌎 
+- A [notebook](https://github.com/aws/amazon-sagemaker-examples/blob/main/introduction_to_amazon_algorithms/jumpstart-foundation-models/text-generation-open-llama.ipynb) on how to deploy Open-LLaMA model for text generation on Amazon SageMaker. 🌎 
 
 ## LlamaConfig
 

docs/source/en/model_doc/llama2.md

@@ -78,6 +78,10 @@ A list of official Hugging Face and community (indicated by 🌎) resources to h
 - A [notebook](https://colab.research.google.com/drive/1PEQyJO1-f6j0S_XJ8DV50NkpzasXkrzd?usp=sharing) on how to fine-tune Llama 2 in Google Colab using QLoRA and 4-bit precision. 🌎
 - A [notebook](https://colab.research.google.com/drive/134o_cXcMe_lsvl15ZE_4Y75Kstepsntu?usp=sharing) on how to fine-tune the "Llama-v2-7b-guanaco" model with 4-bit QLoRA and generate Q&A datasets from PDFs. 🌎
 
+<PipelineTag pipeline="text-classification"/>
+
+- A [notebook](https://colab.research.google.com/drive/1ggaa2oRFphdBmqIjSEbnb_HGkcIRC2ZB?usp=sharing) on how to fine-tune the Llama 2 model with QLoRa, TRL, and Korean text classification dataset. 🌎🇰🇷
+
 ⚗️ Optimization
 - [Fine-tune Llama 2 with DPO](https://huggingface.co/blog/dpo-trl), a guide to using the TRL library's DPO method to fine tune Llama 2 on a specific dataset.
 - [Extended Guide: Instruction-tune Llama 2](https://www.philschmid.de/instruction-tune-llama-2), a guide to training Llama 2 to generate instructions from inputs, transforming the model from instruction-following to instruction-giving.

docs/source/en/model_doc/mms.md

@@ -165,7 +165,147 @@ To further improve performance from ASR models, language model decoding can be u
 
 ### Speech Synthesis (TTS)
 
-Individual TTS models are available for each of the 1100+ languages. The models and inference documentation can be found [here](https://huggingface.co/facebook/mms-tts).
+MMS-TTS uses the same model architecture as VITS, which was added to 🤗 Transformers in v4.33. MMS trains a separate 
+model checkpoint for each of the 1100+ languages in the project. All available checkpoints can be found on the Hugging 
+Face Hub: [facebook/mms-tts](https://huggingface.co/models?sort=trending&search=facebook%2Fmms-tts), and the inference 
+documentation under [VITS](https://huggingface.co/docs/transformers/main/en/model_doc/vits).
+
+#### Inference
+
+To use the MMS model, first update to the latest version of the Transformers library:
+
+```bash
+pip install --upgrade transformers accelerate
+```
+
+Since the flow-based model in VITS is non-deterministic, it is good practice to set a seed to ensure reproducibility of 
+the outputs. 
+
+- For languages with a Roman alphabet, such as English or French, the tokenizer can be used directly to 
+pre-process the text inputs. The following code example runs a forward pass using the MMS-TTS English checkpoint:
+
+```python
+import torch
+from transformers import VitsTokenizer, VitsModel, set_seed
+
+tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-eng")
+model = VitsModel.from_pretrained("facebook/mms-tts-eng")
+
+inputs = tokenizer(text="Hello - my dog is cute", return_tensors="pt")
+
+set_seed(555)  # make deterministic
+
+with torch.no_grad():
+   outputs = model(**inputs)
+
+waveform = outputs.waveform[0]
+```
+
+The resulting waveform can be saved as a `.wav` file:
+
+```python
+import scipy
+
+scipy.io.wavfile.write("synthesized_speech.wav", rate=model.config.sampling_rate, data=waveform)
+```
+
+Or displayed in a Jupyter Notebook / Google Colab:
+
+```python
+from IPython.display import Audio
+
+Audio(waveform, rate=model.config.sampling_rate)
+```
+
+For certain languages with non-Roman alphabets, such as Arabic, Mandarin or Hindi, the [`uroman`](https://github.com/isi-nlp/uroman) 
+perl package is required to pre-process the text inputs to the Roman alphabet.
+
+You can check whether you require the `uroman` package for your language by inspecting the `is_uroman` attribute of 
+the pre-trained `tokenizer`:
+
+```python
+from transformers import VitsTokenizer
+
+tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-eng")
+print(tokenizer.is_uroman)
+```
+
+If required, you should apply the uroman package to your text inputs **prior** to passing them to the `VitsTokenizer`, 
+since currently the tokenizer does not support performing the pre-processing itself.
+
+To do this, first clone the uroman repository to your local machine and set the bash variable `UROMAN` to the local path:
+
+```bash
+git clone https://github.com/isi-nlp/uroman.git
+cd uroman
+export UROMAN=$(pwd)
+```
+
+You can then pre-process the text input using the following code snippet. You can either rely on using the bash variable 
+`UROMAN` to point to the uroman repository, or you can pass the uroman directory as an argument to the `uromaize` function:
+
+```python
+import torch
+from transformers import VitsTokenizer, VitsModel, set_seed
+import os
+import subprocess
+
+tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-kor")
+model = VitsModel.from_pretrained("facebook/mms-tts-kor")
+
+def uromanize(input_string, uroman_path):
+    """Convert non-Roman strings to Roman using the `uroman` perl package."""
+    script_path = os.path.join(uroman_path, "bin", "uroman.pl")
+
+    command = ["perl", script_path]
+
+    process = subprocess.Popen(command, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    # Execute the perl command
+    stdout, stderr = process.communicate(input=input_string.encode())
+
+    if process.returncode != 0:
+        raise ValueError(f"Error {process.returncode}: {stderr.decode()}")
+
+    # Return the output as a string and skip the new-line character at the end
+    return stdout.decode()[:-1]
+
+text = "이봐 무슨 일이야"
+uromaized_text = uromanize(text, uroman_path=os.environ["UROMAN"])
+
+inputs = tokenizer(text=uromaized_text, return_tensors="pt")
+
+set_seed(555)  # make deterministic
+with torch.no_grad():
+   outputs = model(inputs["input_ids"])
+
+waveform = outputs.waveform[0]
+```
+
+**Tips:**
+
+* The MMS-TTS checkpoints are trained on lower-cased, un-punctuated text. By default, the `VitsTokenizer` *normalizes* the inputs by removing any casing and punctuation, to avoid passing out-of-vocabulary characters to the model. Hence, the model is agnostic to casing and punctuation, so these should be avoided in the text prompt. You can disable normalisation by setting `noramlize=False` in the call to the tokenizer, but this will lead to un-expected behaviour and is discouraged.
+* The speaking rate can be varied by setting the attribute `model.speaking_rate` to a chosen value. Likewise, the randomness of the noise is controlled by `model.noise_scale`:
+
+```python
+import torch
+from transformers import VitsTokenizer, VitsModel, set_seed
+
+tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-eng")
+model = VitsModel.from_pretrained("facebook/mms-tts-eng")
+
+inputs = tokenizer(text="Hello - my dog is cute", return_tensors="pt")
+
+# make deterministic
+set_seed(555)  
+
+# make speech faster and more noisy
+model.speaking_rate = 1.5
+model.noise_scale = 0.8
+
+with torch.no_grad():
+   outputs = model(**inputs)
+```
+
 
 ### Language Identification (LID)
 
@@ -173,11 +313,12 @@ Different LID models are available based on the number of languages they can rec
 
 #### Inference
 First, we install transformers and some other libraries
-```
-pip install torch accelerate torchaudio datasets
+
+```bash
+pip install torch accelerate datasets[audio]
 pip install --upgrade transformers
 ````
-pip install torch datasets[audio]
+
 Next, we load a couple of audio samples via `datasets`. Make sure that the audio data is sampled to 16000 kHz.
 
 ```py

docs/source/en/model_doc/musicgen.md

@@ -214,28 +214,7 @@ The MusicGen model can be de-composed into three distinct stages:
 
 Thus, the MusicGen model can either be used as a standalone decoder model, corresponding to the class [`MusicgenForCausalLM`],
 or as a composite model that includes the text encoder and audio encoder/decoder, corresponding to the class
-[`MusicgenForConditionalGeneration`].
-
-Since the text encoder and audio encoder/decoder models are frozen during training, the MusicGen decoder [`MusicgenForCausalLM`]
-can be trained standalone on a dataset of encoder hidden-states and audio codes. For inference, the trained decoder can
-be combined with the frozen text encoder and audio encoder/decoders to recover the composite [`MusicgenForConditionalGeneration`]
-model.
-
-Below, we demonstrate how to construct the composite [`MusicgenForConditionalGeneration`] model from its three constituent
-parts, as would typically be done following training of the MusicGen decoder LM:
-
-```python
->>> from transformers import AutoConfig, AutoModelForTextEncoding, AutoModel, MusicgenForCausalLM, MusicgenForConditionalGeneration
-
->>> text_encoder = AutoModelForTextEncoding.from_pretrained("t5-base")
->>> audio_encoder = AutoModel.from_pretrained("facebook/encodec_32khz")
->>> decoder_config = AutoConfig.from_pretrained("facebook/musicgen-small").decoder
->>> decoder = MusicgenForCausalLM.from_pretrained("facebook/musicgen-small", **decoder_config)
-
->>> model = MusicgenForConditionalGeneration.from_sub_models_pretrained(text_encoder, audio_encoder, decoder)
-```
-
-If only the decoder needs to be loaded from the pre-trained checkpoint for the composite model, it can be loaded by first 
+[`MusicgenForConditionalGeneration`]. If only the decoder needs to be loaded from the pre-trained checkpoint, it can be loaded by first 
 specifying the correct config, or be accessed through the `.decoder` attribute of the composite model:
 
 ```python
@@ -249,6 +228,11 @@ specifying the correct config, or be accessed through the `.decoder` attribute o
 >>> decoder = MusicgenForConditionalGeneration.from_pretrained("facebook/musicgen-small").decoder
 ```
 
+Since the text encoder and audio encoder/decoder models are frozen during training, the MusicGen decoder [`MusicgenForCausalLM`]
+can be trained standalone on a dataset of encoder hidden-states and audio codes. For inference, the trained decoder can
+be combined with the frozen text encoder and audio encoder/decoders to recover the composite [`MusicgenForConditionalGeneration`]
+model.
+
 Tips:
 * MusicGen is trained on the 32kHz checkpoint of Encodec. You should ensure you use a compatible version of the Encodec model.
 * Sampling mode tends to deliver better results than greedy - you can toggle sampling with the variable `do_sample` in the call to [`MusicgenForConditionalGeneration.generate`]

docs/source/en/model_doc/persimmon.md

@@ -0,0 +1,96 @@
+<!--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.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+# Persimmon
+
+## Overview
+
+The Persimmon model was created by [ADEPT](https://www.adept.ai/blog/persimmon-8b), and authored by Erich Elsen, Augustus Odena, Maxwell Nye, Sağnak Taşırlar, Tri Dao, Curtis Hawthorne, Deepak Moparthi, Arushi Somani.
+
+The authors introduced Persimmon-8B, a decoder model based on the classic transformers architecture, with query and key normalization. Persimmon-8B is a fully permissively-licensed model with approximately 8 billion parameters, released under the Apache license.  Some of the key attributes of Persimmon-8B are long context size (16K), performance, and capabilities for multimodal extensions.
+
+The authors showcase their approach to model evaluation, focusing on practical text generation, mirroring how users interact with language models. The work also includes a comparative analysis, pitting Persimmon-8B against other prominent models (MPT 7B Instruct and Llama 2 Base 7B 1-Shot), across various evaluation tasks. The results demonstrate Persimmon-8B's competitive performance, even with limited training data.
+
+In terms of model details, the work outlines the architecture and training methodology of Persimmon-8B, providing insights into its design choices, sequence length, and dataset composition. The authors present a fast inference code that outperforms traditional implementations through operator fusion and CUDA graph utilization while maintaining code coherence. They express their anticipation of how the community will leverage this contribution to drive innovation, hinting at further upcoming releases as part of an ongoing series of developments.
+
+
+<Tip warning={true}>
+
+The `Persimmon` models were trained using `bfloat16`, but the original inference uses `float16` The checkpoints uploaded on the hub use `torch_dtype = 'float16'` which will be
+used by the `AutoModel` API to cast the checkpoints from `torch.float32` to `torch.float16`. 
+
+The `dtype` of the online weights is mostly irrelevant, unless you are using `torch_dtype="auto"` when initializing a model using `model = AutoModelForCausalLM.from_pretrained("path", torch_dtype = "auto")`. The reason is that the model will first be downloaded ( using the `dtype` of the checkpoints online) then it will be cast to the default `dtype` of `torch` (becomes `torch.float32`). Users should specify the `torch_dtype` they want, and if they don't it will be `torch.float32`.
+
+Finetuning the model in `float16` is not recommended and known to produce `nan`, as such the model should be fine-tuned in `bfloat16`.
+
+</Tip>
+
+
+Tips:
+
+- To convert the model, you need to clone the original repository using `git clone https://github.com/persimmon-ai-labs/adept-inference`, then get the checkpoints:
+
+```bash
+git clone https://github.com/persimmon-ai-labs/adept-inference
+wget https://axtkn4xl5cip.objectstorage.us-phoenix-1.oci.customer-oci.com/n/axtkn4xl5cip/b/adept-public-data/o/8b_base_model_release.tar
+tar -xvf 8b_base_model_release.tar
+python src/transformers/models/persimmon/convert_persimmon_weights_to_hf.py  --input_dir /path/to/downloaded/persimmon/weights/ --output_dir /output/path \
+    --pt_model_path /path/to/8b_chat_model_release/iter_0001251/mp_rank_00/model_optim_rng.pt
+    --ada_lib_path /path/to/adept-inference
+```
+
+For the chat model:
+```bash
+wget https://axtkn4xl5cip.objectstorage.us-phoenix-1.oci.customer-oci.com/n/axtkn4xl5cip/b/adept-public-data/o/8b_chat_model_release.tar
+tar -xvf 8b_base_model_release.tar
+```
+
+Thereafter, models can be loaded via:
+
+```py
+from transformers import PersimmonForCausalLM, PersimmonTokenizer
+
+model = PersimmonForCausalLM.from_pretrained("/output/path")
+tokenizer = PersimmonTokenizer.from_pretrained("/output/path")
+```
+
+This model was contributed by [ArthurZ](https://huggingface.co/ArthurZ).
+The original code can be found [here](https://github.com/persimmon-ai-labs/adept-inference).
+
+- Perismmon uses a `sentencepiece` based tokenizer, with a `Unigram` model. It supports bytefallback, which is only available in `tokenizers==0.14.0` for the fast tokenizer.
+The `LlamaTokenizer` is used as it is a standard wrapper around sentencepiece. The `chat` template will be updated with the templating functions in a follow up PR!
+
+- The authors suggest to use the following prompt format for the chat mode: `f"human: {prompt}\n\nadept:"`
+
+
+## PersimmonConfig
+
+[[autodoc]] PersimmonConfig
+
+## PersimmonModel
+
+[[autodoc]] PersimmonModel
+    - forward
+
+## PersimmonForCausalLM
+
+[[autodoc]] PersimmonForCausalLM
+    - forward
+
+## PersimmonForSequenceClassification
+
+[[autodoc]] PersimmonForSequenceClassification
+    - forward

docs/source/en/model_doc/pop2piano.md

@@ -12,6 +12,12 @@ specific language governing permissions and limitations under the License.
 
 # Pop2Piano
 
+<div class="flex flex-wrap space-x-1">
+<a href="https://huggingface.co/spaces/sweetcocoa/pop2piano">
+<img alt="Spaces" src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue">
+</a>
+</div>
+
 ## Overview
 
 The Pop2Piano model was proposed in [Pop2Piano : Pop Audio-based Piano Cover Generation](https://arxiv.org/abs/2211.00895) by Jongho Choi and Kyogu Lee.

docs/source/en/model_doc/vits.md

@@ -0,0 +1,162 @@
+<!--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.
+-->
+
+# VITS
+
+## Overview
+
+The VITS model was proposed in [Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech](https://arxiv.org/abs/2106.06103) by Jaehyeon Kim, Jungil Kong, Juhee Son.
+
+
+VITS (**V**ariational **I**nference with adversarial learning for end-to-end **T**ext-to-**S**peech) is an end-to-end 
+speech synthesis model that predicts a speech waveform conditional on an input text sequence. It is a conditional variational 
+autoencoder (VAE) comprised of a posterior encoder, decoder, and conditional prior.
+
+A set of spectrogram-based acoustic features are predicted by the flow-based module, which is formed of a Transformer-based
+text encoder and multiple coupling layers. The spectrogram is decoded using a stack of transposed convolutional layers,
+much in the same style as the HiFi-GAN vocoder. Motivated by the one-to-many nature of the TTS problem, where the same text 
+input can be spoken in multiple ways, the model also includes a stochastic duration predictor, which allows the model to 
+synthesise speech with different rhythms from the same input text. 
+
+The model is trained end-to-end with a combination of losses derived from variational lower bound and adversarial training. 
+To improve the expressiveness of the model, normalizing flows are applied to the conditional prior distribution. During 
+inference, the text encodings are up-sampled based on the duration prediction module, and then mapped into the 
+waveform using a cascade of the flow module and HiFi-GAN decoder. Due to the stochastic nature of the duration predictor,
+the model is non-deterministic, and thus requires a fixed seed to generate the same speech waveform.
+
+The abstract from the paper is the following:
+
+*Several recent end-to-end text-to-speech (TTS) models enabling single-stage training and parallel sampling have been proposed, but their sample quality does not match that of two-stage TTS systems. In this work, we present a parallel end-to-end TTS method that generates more natural sounding audio than current two-stage models. Our method adopts variational inference augmented with normalizing flows and an adversarial training process, which improves the expressive power of generative modeling. We also propose a stochastic duration predictor to synthesize speech with diverse rhythms from input text. With the uncertainty modeling over latent variables and the stochastic duration predictor, our method expresses the natural one-to-many relationship in which a text input can be spoken in multiple ways with different pitches and rhythms. A subjective human evaluation (mean opinion score, or MOS) on the LJ Speech, a single speaker dataset, shows that our method outperforms the best publicly available TTS systems and achieves a MOS comparable to ground truth.*
+
+This model can also be used with TTS checkpoints from [Massively Multilingual Speech (MMS)](https://arxiv.org/abs/2305.13516) 
+as these checkpoints use the same architecture and a slightly modified tokenizer.
+
+This model was contributed by [Matthijs](https://huggingface.co/Matthijs) and [sanchit-gandhi](https://huggingface.co/sanchit-gandhi). The original code can be found [here](https://github.com/jaywalnut310/vits).
+
+## Model Usage
+
+Both the VITS and MMS-TTS checkpoints can be used with the same API. Since the flow-based model is non-deterministic, it 
+is good practice to set a seed to ensure reproducibility of the outputs. For languages with a Roman alphabet, 
+such as English or French, the tokenizer can be used directly to pre-process the text inputs. The following code example 
+runs a forward pass using the MMS-TTS English checkpoint:
+
+```python
+import torch
+from transformers import VitsTokenizer, VitsModel, set_seed
+
+tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-eng")
+model = VitsModel.from_pretrained("facebook/mms-tts-eng")
+
+inputs = tokenizer(text="Hello - my dog is cute", return_tensors="pt")
+
+set_seed(555)  # make deterministic
+
+with torch.no_grad():
+   outputs = model(**inputs)
+
+waveform = outputs.waveform[0]
+```
+
+The resulting waveform can be saved as a `.wav` file:
+
+```python
+import scipy
+
+scipy.io.wavfile.write("techno.wav", rate=model.config.sampling_rate, data=waveform)
+```
+
+Or displayed in a Jupyter Notebook / Google Colab:
+
+```python
+from IPython.display import Audio
+
+Audio(waveform, rate=model.config.sampling_rate)
+```
+
+For certain languages with a non-Roman alphabet, such as Arabic, Mandarin or Hindi, the [`uroman`](https://github.com/isi-nlp/uroman) 
+perl package is required to pre-process the text inputs to the Roman alphabet.
+
+You can check whether you require the `uroman` package for your language by inspecting the `is_uroman` attribute of 
+the pre-trained `tokenizer`:
+
+```python
+from transformers import VitsTokenizer
+
+tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-eng")
+print(tokenizer.is_uroman)
+```
+
+If required, you should apply the uroman package to your text inputs **prior** to passing them to the `VitsTokenizer`, 
+since currently the tokenizer does not support performing the pre-processing itself.  
+
+To do this, first clone the uroman repository to your local machine and set the bash variable `UROMAN` to the local path:
+
+```bash
+git clone https://github.com/isi-nlp/uroman.git
+cd uroman
+export UROMAN=$(pwd)
+```
+
+You can then pre-process the text input using the following code snippet. You can either rely on using the bash variable 
+`UROMAN` to point to the uroman repository, or you can pass the uroman directory as an argument to the `uromaize` function:
+
+```python
+import torch
+from transformers import VitsTokenizer, VitsModel, set_seed
+import os
+import subprocess
+
+tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-kor")
+model = VitsModel.from_pretrained("facebook/mms-tts-kor")
+
+def uromanize(input_string, uroman_path):
+    """Convert non-Roman strings to Roman using the `uroman` perl package."""
+    script_path = os.path.join(uroman_path, "bin", "uroman.pl")
+
+    command = ["perl", script_path]
+
+    process = subprocess.Popen(command, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    # Execute the perl command
+    stdout, stderr = process.communicate(input=input_string.encode())
+
+    if process.returncode != 0:
+        raise ValueError(f"Error {process.returncode}: {stderr.decode()}")
+
+    # Return the output as a string and skip the new-line character at the end
+    return stdout.decode()[:-1]
+
+text = "이봐 무슨 일이야"
+uromaized_text = uromanize(text, uroman_path=os.environ["UROMAN"])
+
+inputs = tokenizer(text=uromaized_text, return_tensors="pt")
+
+set_seed(555)  # make deterministic
+with torch.no_grad():
+   outputs = model(inputs["input_ids"])
+
+waveform = outputs.waveform[0]
+```
+
+## VitsConfig
+
+[[autodoc]] VitsConfig
+
+## VitsTokenizer
+
+[[autodoc]] VitsTokenizer
+    - __call__
+    - save_vocabulary
+
+## VitsModel
+
+[[autodoc]] VitsModel
+    - forward

docs/source/en/model_memory_anatomy.md

@@ -76,7 +76,7 @@ GPU memory occupied: 0 MB.
 
 That looks good: the GPU memory is not occupied as we would expect before we load any models. If that's not the case on 
 your machine make sure to stop all processes that are using GPU memory. However, not all free GPU memory can be used by 
-the user. When a model is loaded to the GPU also the kernels are loaded which can take up 1-2GB of memory. To see how 
+the user. When a model is loaded to the GPU the kernels are also loaded, which can take up 1-2GB of memory. To see how 
 much it is we load a tiny tensor into the GPU which triggers the kernels to be loaded as well.
 
 ```py
@@ -105,7 +105,7 @@ how much space just the weights use.
 GPU memory occupied: 2631 MB.
 ```
 
-We can see that the model weights alone take up 1.3 GB of the GPU memory. The exact number depends on the specific 
+We can see that the model weights alone take up 1.3 GB of GPU memory. The exact number depends on the specific 
 GPU you are using. Note that on newer GPUs a model can sometimes take up more space since the weights are loaded in an 
 optimized fashion that speeds up the usage of the model. Now we can also quickly check if we get the same result 
 as with `nvidia-smi` CLI:
@@ -184,7 +184,7 @@ GPU memory occupied: 14949 MB.
 We see that already a relatively small batch size almost fills up our GPU's entire memory. However, a larger batch size 
 can often result in faster model convergence or better end performance. So ideally we want to tune the batch size to our
 model's needs and not to the GPU limitations. What's interesting is that we use much more memory than the size of the model. 
-To understand a bit better why this is the case let's have look at a model's operations and memory needs.
+To understand a bit better why this is the case let's have a look at a model's operations and memory needs.
 
 ## Anatomy of Model's Operations
 

docs/source/en/model_sharing.md

@@ -95,7 +95,7 @@ Specify `from_pt=True` to convert a checkpoint from PyTorch to TensorFlow:
 >>> tf_model = TFDistilBertForSequenceClassification.from_pretrained("path/to/awesome-name-you-picked", from_pt=True)
 ```
 
-Then you can save your new TensorFlow model with it's new checkpoint:
+Then you can save your new TensorFlow model with its new checkpoint:
 
 ```py
 >>> tf_model.save_pretrained("path/to/awesome-name-you-picked")
@@ -201,7 +201,7 @@ Or perhaps you'd like to add the TensorFlow version of your fine-tuned PyTorch m
 >>> tf_model.push_to_hub("my-awesome-model")
 ```
 
-Now when you navigate to the your Hugging Face profile, you should see your newly created model repository. Clicking on the **Files** tab will display all the files you've uploaded to the repository.
+Now when you navigate to your Hugging Face profile, you should see your newly created model repository. Clicking on the **Files** tab will display all the files you've uploaded to the repository.
 
 For more details on how to create and upload files to a repository, refer to the Hub documentation [here](https://huggingface.co/docs/hub/how-to-upstream).
 

docs/source/en/perf_train_gpu_one.md

@@ -237,10 +237,11 @@ For example if you have [NVIDIA/apex](https://github.com/NVIDIA/apex) installed,
 fastest training experience among all supported AdamW optimizers.
 
 [`Trainer`] integrates a variety of optimizers that can be used out of box: `adamw_hf`, `adamw_torch`, `adamw_torch_fused`, 
-`adamw_apex_fused`, `adamw_anyprecision` or `adafactor`. More optimizers can be plugged in via a third-party implementation.
+`adamw_apex_fused`, `adamw_anyprecision`, `adafactor`, or `adamw_bnb_8bit`. More optimizers can be plugged in via a third-party implementation.
 
-Let's take a closer look at two alternatives to AdamW optimizer - Adafactor (available in Trainer), and 8bit BNB quantized 
-optimizer (third-party implementation).
+Let's take a closer look at two alternatives to AdamW optimizer:
+1. `adafactor` which is available in [`Trainer`]
+2. `adamw_bnb_8bit` is also available in Trainer, but a third-party integration is provided below for demonstration.
 
 For comparison, for a 3B-parameter model, like “t5-3b”: 
 * A standard AdamW optimizer will need 24GB of GPU memory because it uses 8 bytes for each parameter (8*3 => 24GB)
@@ -269,7 +270,13 @@ Instead of aggregating optimizer states like Adafactor, 8-bit Adam keeps the ful
 means that it stores the state with lower precision and dequantizes it only for the optimization. This is similar to the 
 idea behind mixed precision training.
 
-To use the 8-bit optimizer, you need to install it separately and then pass it as a custom optimizer to the [`Trainer`]. 
+To use `adamw_bnb_8bit`, you simply need to set `optim="adamw_bnb_8bit"` in [`TrainingArguments`]:
+
+```py
+training_args = TrainingArguments(per_device_train_batch_size=4, optim="adamw_bnb_8bit", **default_args)
+```
+
+However, we can also use a third-party implementation of the 8-bit optimizer for demonstration purposes to see how that can be integrated.
 
 First, follow the installation guide in the GitHub [repo](https://github.com/TimDettmers/bitsandbytes) to install the `bitsandbytes` library 
 that implements the 8-bit Adam optimizer.
@@ -311,13 +318,6 @@ adam_bnb_optim = bnb.optim.Adam8bit(
 )
 ```
 
-<Tip>
-
-To use the 8-bit optimizer with an existing pretrained model, you need to make a change to the embedding layer.
-Read [this issue](https://github.com/huggingface/transformers/issues/14819) for more information.
-
-</Tip>
-
 Finally, pass the custom optimizer as an argument to the `Trainer`:
 
 ```py

docs/source/en/tasks/language_modeling.md

@@ -37,7 +37,7 @@ You can finetune other architectures for causal language modeling following the
 Choose one of the following architectures:
 
 <!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->
-[BART](../model_doc/bart), [BERT](../model_doc/bert), [Bert Generation](../model_doc/bert-generation), [BigBird](../model_doc/big_bird), [BigBird-Pegasus](../model_doc/bigbird_pegasus), [BioGpt](../model_doc/biogpt), [Blenderbot](../model_doc/blenderbot), [BlenderbotSmall](../model_doc/blenderbot-small), [BLOOM](../model_doc/bloom), [CamemBERT](../model_doc/camembert), [CodeLlama](../model_doc/code_llama), [CodeGen](../model_doc/codegen), [CPM-Ant](../model_doc/cpmant), [CTRL](../model_doc/ctrl), [Data2VecText](../model_doc/data2vec-text), [ELECTRA](../model_doc/electra), [ERNIE](../model_doc/ernie), [Falcon](../model_doc/falcon), [GIT](../model_doc/git), [GPT-Sw3](../model_doc/gpt-sw3), [OpenAI GPT-2](../model_doc/gpt2), [GPTBigCode](../model_doc/gpt_bigcode), [GPT Neo](../model_doc/gpt_neo), [GPT NeoX](../model_doc/gpt_neox), [GPT NeoX Japanese](../model_doc/gpt_neox_japanese), [GPT-J](../model_doc/gptj), [LLaMA](../model_doc/llama), [Marian](../model_doc/marian), [mBART](../model_doc/mbart), [MEGA](../model_doc/mega), [Megatron-BERT](../model_doc/megatron-bert), [MPT](../model_doc/mpt), [MusicGen](../model_doc/musicgen), [MVP](../model_doc/mvp), [OpenLlama](../model_doc/open-llama), [OpenAI GPT](../model_doc/openai-gpt), [OPT](../model_doc/opt), [Pegasus](../model_doc/pegasus), [PLBart](../model_doc/plbart), [ProphetNet](../model_doc/prophetnet), [QDQBert](../model_doc/qdqbert), [Reformer](../model_doc/reformer), [RemBERT](../model_doc/rembert), [RoBERTa](../model_doc/roberta), [RoBERTa-PreLayerNorm](../model_doc/roberta-prelayernorm), [RoCBert](../model_doc/roc_bert), [RoFormer](../model_doc/roformer), [RWKV](../model_doc/rwkv), [Speech2Text2](../model_doc/speech_to_text_2), [Transformer-XL](../model_doc/transfo-xl), [TrOCR](../model_doc/trocr), [XGLM](../model_doc/xglm), [XLM](../model_doc/xlm), [XLM-ProphetNet](../model_doc/xlm-prophetnet), [XLM-RoBERTa](../model_doc/xlm-roberta), [XLM-RoBERTa-XL](../model_doc/xlm-roberta-xl), [XLNet](../model_doc/xlnet), [X-MOD](../model_doc/xmod)
+[BART](../model_doc/bart), [BERT](../model_doc/bert), [Bert Generation](../model_doc/bert-generation), [BigBird](../model_doc/big_bird), [BigBird-Pegasus](../model_doc/bigbird_pegasus), [BioGpt](../model_doc/biogpt), [Blenderbot](../model_doc/blenderbot), [BlenderbotSmall](../model_doc/blenderbot-small), [BLOOM](../model_doc/bloom), [CamemBERT](../model_doc/camembert), [CodeLlama](../model_doc/code_llama), [CodeGen](../model_doc/codegen), [CPM-Ant](../model_doc/cpmant), [CTRL](../model_doc/ctrl), [Data2VecText](../model_doc/data2vec-text), [ELECTRA](../model_doc/electra), [ERNIE](../model_doc/ernie), [Falcon](../model_doc/falcon), [GIT](../model_doc/git), [GPT-Sw3](../model_doc/gpt-sw3), [OpenAI GPT-2](../model_doc/gpt2), [GPTBigCode](../model_doc/gpt_bigcode), [GPT Neo](../model_doc/gpt_neo), [GPT NeoX](../model_doc/gpt_neox), [GPT NeoX Japanese](../model_doc/gpt_neox_japanese), [GPT-J](../model_doc/gptj), [LLaMA](../model_doc/llama), [Marian](../model_doc/marian), [mBART](../model_doc/mbart), [MEGA](../model_doc/mega), [Megatron-BERT](../model_doc/megatron-bert), [MPT](../model_doc/mpt), [MusicGen](../model_doc/musicgen), [MVP](../model_doc/mvp), [OpenLlama](../model_doc/open-llama), [OpenAI GPT](../model_doc/openai-gpt), [OPT](../model_doc/opt), [Pegasus](../model_doc/pegasus), [Persimmon](../model_doc/persimmon), [PLBart](../model_doc/plbart), [ProphetNet](../model_doc/prophetnet), [QDQBert](../model_doc/qdqbert), [Reformer](../model_doc/reformer), [RemBERT](../model_doc/rembert), [RoBERTa](../model_doc/roberta), [RoBERTa-PreLayerNorm](../model_doc/roberta-prelayernorm), [RoCBert](../model_doc/roc_bert), [RoFormer](../model_doc/roformer), [RWKV](../model_doc/rwkv), [Speech2Text2](../model_doc/speech_to_text_2), [Transformer-XL](../model_doc/transfo-xl), [TrOCR](../model_doc/trocr), [XGLM](../model_doc/xglm), [XLM](../model_doc/xlm), [XLM-ProphetNet](../model_doc/xlm-prophetnet), [XLM-RoBERTa](../model_doc/xlm-roberta), [XLM-RoBERTa-XL](../model_doc/xlm-roberta-xl), [XLNet](../model_doc/xlnet), [X-MOD](../model_doc/xmod)
 
 
 

docs/source/en/tasks/sequence_classification.md

@@ -33,7 +33,7 @@ The task illustrated in this tutorial is supported by the following model archit
 <!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->
 
 
-[ALBERT](../model_doc/albert), [BART](../model_doc/bart), [BERT](../model_doc/bert), [BigBird](../model_doc/big_bird), [BigBird-Pegasus](../model_doc/bigbird_pegasus), [BioGpt](../model_doc/biogpt), [BLOOM](../model_doc/bloom), [CamemBERT](../model_doc/camembert), [CANINE](../model_doc/canine), [CodeLlama](../model_doc/code_llama), [ConvBERT](../model_doc/convbert), [CTRL](../model_doc/ctrl), [Data2VecText](../model_doc/data2vec-text), [DeBERTa](../model_doc/deberta), [DeBERTa-v2](../model_doc/deberta-v2), [DistilBERT](../model_doc/distilbert), [ELECTRA](../model_doc/electra), [ERNIE](../model_doc/ernie), [ErnieM](../model_doc/ernie_m), [ESM](../model_doc/esm), [Falcon](../model_doc/falcon), [FlauBERT](../model_doc/flaubert), [FNet](../model_doc/fnet), [Funnel Transformer](../model_doc/funnel), [GPT-Sw3](../model_doc/gpt-sw3), [OpenAI GPT-2](../model_doc/gpt2), [GPTBigCode](../model_doc/gpt_bigcode), [GPT Neo](../model_doc/gpt_neo), [GPT NeoX](../model_doc/gpt_neox), [GPT-J](../model_doc/gptj), [I-BERT](../model_doc/ibert), [LayoutLM](../model_doc/layoutlm), [LayoutLMv2](../model_doc/layoutlmv2), [LayoutLMv3](../model_doc/layoutlmv3), [LED](../model_doc/led), [LiLT](../model_doc/lilt), [LLaMA](../model_doc/llama), [Longformer](../model_doc/longformer), [LUKE](../model_doc/luke), [MarkupLM](../model_doc/markuplm), [mBART](../model_doc/mbart), [MEGA](../model_doc/mega), [Megatron-BERT](../model_doc/megatron-bert), [MobileBERT](../model_doc/mobilebert), [MPNet](../model_doc/mpnet), [MPT](../model_doc/mpt), [MRA](../model_doc/mra), [MT5](../model_doc/mt5), [MVP](../model_doc/mvp), [Nezha](../model_doc/nezha), [Nyströmformer](../model_doc/nystromformer), [OpenLlama](../model_doc/open-llama), [OpenAI GPT](../model_doc/openai-gpt), [OPT](../model_doc/opt), [Perceiver](../model_doc/perceiver), [PLBart](../model_doc/plbart), [QDQBert](../model_doc/qdqbert), [Reformer](../model_doc/reformer), [RemBERT](../model_doc/rembert), [RoBERTa](../model_doc/roberta), [RoBERTa-PreLayerNorm](../model_doc/roberta-prelayernorm), [RoCBert](../model_doc/roc_bert), [RoFormer](../model_doc/roformer), [SqueezeBERT](../model_doc/squeezebert), [T5](../model_doc/t5), [TAPAS](../model_doc/tapas), [Transformer-XL](../model_doc/transfo-xl), [UMT5](../model_doc/umt5), [XLM](../model_doc/xlm), [XLM-RoBERTa](../model_doc/xlm-roberta), [XLM-RoBERTa-XL](../model_doc/xlm-roberta-xl), [XLNet](../model_doc/xlnet), [X-MOD](../model_doc/xmod), [YOSO](../model_doc/yoso)
+[ALBERT](../model_doc/albert), [BART](../model_doc/bart), [BERT](../model_doc/bert), [BigBird](../model_doc/big_bird), [BigBird-Pegasus](../model_doc/bigbird_pegasus), [BioGpt](../model_doc/biogpt), [BLOOM](../model_doc/bloom), [CamemBERT](../model_doc/camembert), [CANINE](../model_doc/canine), [CodeLlama](../model_doc/code_llama), [ConvBERT](../model_doc/convbert), [CTRL](../model_doc/ctrl), [Data2VecText](../model_doc/data2vec-text), [DeBERTa](../model_doc/deberta), [DeBERTa-v2](../model_doc/deberta-v2), [DistilBERT](../model_doc/distilbert), [ELECTRA](../model_doc/electra), [ERNIE](../model_doc/ernie), [ErnieM](../model_doc/ernie_m), [ESM](../model_doc/esm), [Falcon](../model_doc/falcon), [FlauBERT](../model_doc/flaubert), [FNet](../model_doc/fnet), [Funnel Transformer](../model_doc/funnel), [GPT-Sw3](../model_doc/gpt-sw3), [OpenAI GPT-2](../model_doc/gpt2), [GPTBigCode](../model_doc/gpt_bigcode), [GPT Neo](../model_doc/gpt_neo), [GPT NeoX](../model_doc/gpt_neox), [GPT-J](../model_doc/gptj), [I-BERT](../model_doc/ibert), [LayoutLM](../model_doc/layoutlm), [LayoutLMv2](../model_doc/layoutlmv2), [LayoutLMv3](../model_doc/layoutlmv3), [LED](../model_doc/led), [LiLT](../model_doc/lilt), [LLaMA](../model_doc/llama), [Longformer](../model_doc/longformer), [LUKE](../model_doc/luke), [MarkupLM](../model_doc/markuplm), [mBART](../model_doc/mbart), [MEGA](../model_doc/mega), [Megatron-BERT](../model_doc/megatron-bert), [MobileBERT](../model_doc/mobilebert), [MPNet](../model_doc/mpnet), [MPT](../model_doc/mpt), [MRA](../model_doc/mra), [MT5](../model_doc/mt5), [MVP](../model_doc/mvp), [Nezha](../model_doc/nezha), [Nyströmformer](../model_doc/nystromformer), [OpenLlama](../model_doc/open-llama), [OpenAI GPT](../model_doc/openai-gpt), [OPT](../model_doc/opt), [Perceiver](../model_doc/perceiver), [Persimmon](../model_doc/persimmon), [PLBart](../model_doc/plbart), [QDQBert](../model_doc/qdqbert), [Reformer](../model_doc/reformer), [RemBERT](../model_doc/rembert), [RoBERTa](../model_doc/roberta), [RoBERTa-PreLayerNorm](../model_doc/roberta-prelayernorm), [RoCBert](../model_doc/roc_bert), [RoFormer](../model_doc/roformer), [SqueezeBERT](../model_doc/squeezebert), [T5](../model_doc/t5), [TAPAS](../model_doc/tapas), [Transformer-XL](../model_doc/transfo-xl), [UMT5](../model_doc/umt5), [XLM](../model_doc/xlm), [XLM-RoBERTa](../model_doc/xlm-roberta), [XLM-RoBERTa-XL](../model_doc/xlm-roberta-xl), [XLNet](../model_doc/xlnet), [X-MOD](../model_doc/xmod), [YOSO](../model_doc/yoso)
 
 
 

docs/source/en/tasks/text-to-speech.md

@@ -19,16 +19,40 @@ rendered properly in your Markdown viewer.
 [[open-in-colab]]
 
 Text-to-speech (TTS) is the task of creating natural-sounding speech from text, where the speech can be generated in multiple 
-languages and for multiple speakers. The only text-to-speech model currently available in 🤗 Transformers 
-is [SpeechT5](model_doc/speecht5), though more will be added in the future. SpeechT5 is pre-trained on a combination of 
+languages and for multiple speakers. Several text-to-speech models are currently available in 🤗 Transformers, such as 
+[Bark](../model_doc/bark), [MMS](../model_doc/mms), [VITS](../model_doc/vits) and [SpeechT5](../model_doc/speecht5). 
+
+You can easily generate audio using the `"text-to-audio"` pipeline (or its alias - `"text-to-speech"`). Some models, like Bark, 
+can also be conditioned to generate non-verbal communications such as laughing, sighing and crying, or even add music.
+Here's an example of how you would use the `"text-to-speech"` pipeline with Bark: 
+
+```py
+>>> from transformers import pipeline
+
+>>> pipe = pipeline("text-to-speech", model="suno/bark-small")
+>>> text = "[clears throat] This is a test ... and I just took a long pause."
+>>> output = pipe(text)
+```
+
+Here's a code snippet you can use to listen to the resulting audio in a notebook: 
+
+```python
+>>> from IPython.display import Audio
+>>> Audio(output["audio"], rate=output["sampling_rate"])
+```
+
+For more examples on what Bark and other pretrained TTS models can do, refer to our 
+[Audio course](https://huggingface.co/learn/audio-course/chapter6/pre-trained_models). 
+
+If you are looking to fine-tune a TTS model, you can currently fine-tune SpeechT5 only. SpeechT5 is pre-trained on a combination of 
 speech-to-text and text-to-speech data, allowing it to learn a unified space of hidden representations shared by both text 
 and speech. This means that the same pre-trained model can be fine-tuned for different tasks. Furthermore, SpeechT5 
 supports multiple speakers through x-vector speaker embeddings. 
 
-This guide illustrates how to:
+The remainder of this guide illustrates how to:
 
-1. Fine-tune [SpeechT5](model_doc/speecht5) that was originally trained on English speech on the Dutch (`nl`) language subset of the [VoxPopuli](https://huggingface.co/datasets/facebook/voxpopuli) dataset.
-2. Use your fine-tuned model for inference.
+1. Fine-tune [SpeechT5](../model_doc/speecht5) that was originally trained on English speech on the Dutch (`nl`) language subset of the [VoxPopuli](https://huggingface.co/datasets/facebook/voxpopuli) dataset.
+2. Use your refined model for inference in one of two ways: using a pipeline or directly.
 
 Before you begin, make sure you have all the necessary libraries installed:
 
@@ -485,6 +509,12 @@ the `per_device_train_batch_size` incrementally by factors of 2 and increase `gr
 >>> trainer.train()
 ```
 
+To be able to use your checkpoint with a pipeline, make sure to save the processor with the checkpoint: 
+
+```py
+>>> processor.save_pretrained("YOUR_ACCOUNT_NAME/speecht5_finetuned_voxpopuli_nl")
+```
+
 Push the final model to the 🤗 Hub:
 
 ```py
@@ -493,29 +523,70 @@ Push the final model to the 🤗 Hub:
 
 ## Inference
 
+### Inference with a pipeline
+
 Great, now that you've fine-tuned a model, you can use it for inference!
-Load the model from the 🤗 Hub (make sure to use your account name in the following code snippet): 
+First, let's see how you can use it with a corresponding pipeline. Let's create a `"text-to-speech"` pipeline with your 
+checkpoint: 
+
+```py
+>>> from transformers import pipeline
+
+>>> pipe = pipeline("text-to-speech", model="YOUR_ACCOUNT_NAME/speecht5_finetuned_voxpopuli_nl")
+```
+
+Pick a piece of text in Dutch you'd like narrated, e.g.:
+
+```py
+>>> text = "hallo allemaal, ik praat nederlands. groetjes aan iedereen!"
+```
+
+To use SpeechT5 with the pipeline, you'll need a speaker embedding. Let's get it from an example in the test dataset: 
+
+```py
+>>> example = dataset["test"][304]
+>>> speaker_embeddings = torch.tensor(example["speaker_embeddings"]).unsqueeze(0)
+```
+
+Now you can pass the text and speaker embeddings to the pipeline, and it will take care of the rest: 
+
+```py
+>>> forward_params = {"speaker_embeddings": speaker_embeddings}
+>>> output = pipe(text, forward_params=forward_params)
+>>> output
+{'audio': array([-6.82714235e-05, -4.26525949e-04,  1.06134125e-04, ...,
+        -1.22392643e-03, -7.76011671e-04,  3.29112721e-04], dtype=float32),
+ 'sampling_rate': 16000}
+```
+
+You can then listen to the result:
+
+```py
+>>> from IPython.display import Audio
+>>> Audio(output['audio'], rate=output['sampling_rate']) 
+```
+
+### Run inference manually
+
+You can achieve the same inference results without using the pipeline, however, more steps will be required. 
+
+Load the model from the 🤗 Hub: 
 
 ```py
 >>> model = SpeechT5ForTextToSpeech.from_pretrained("YOUR_ACCOUNT/speecht5_finetuned_voxpopuli_nl")
 ```
 
-Pick an example, here we'll take one from the test dataset. Obtain a speaker embedding. 
+Pick an example from the test dataset obtain a speaker embedding. 
 
 ```py 
 >>> example = dataset["test"][304]
 >>> speaker_embeddings = torch.tensor(example["speaker_embeddings"]).unsqueeze(0)
 ```
 
-Define some input text and tokenize it.
+Define the input text and tokenize it.
 
 ```py 
 >>> text = "hallo allemaal, ik praat nederlands. groetjes aan iedereen!"
-```
-
-Preprocess the input text: 
-
-```py
 >>> inputs = processor(text=text, return_tensors="pt")
 ```
 

docs/source/en/tasks/token_classification.md

@@ -32,8 +32,7 @@ The task illustrated in this tutorial is supported by the following model archit
 
 <!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->
 
-[ALBERT](../model_doc/albert), [BERT](../model_doc/bert), [BigBird](../model_doc/big_bird), [BioGpt](../model_doc/biogpt), [BLOOM](../model_doc/bloom), [CamemBERT](../model_doc/camembert), [CANINE](../model_doc/canine), [ConvBERT](../model_doc/convbert), [Data2VecText](../model_doc/data2vec-text), [DeBERTa](../model_doc/deberta), [DeBERTa-v2](../model_doc/deberta-v2), [DistilBERT](../model_doc/distilbert), [ELECTRA](../model_doc/electra), [ERNIE](../model_doc/ernie), [ErnieM](../model_doc/ernie_m), [ESM](../model_doc/esm), [Falcon](../model_doc/falcon), [FlauBERT](../model_doc/flaubert), [FNet](../model_doc/fnet), [Funnel Transformer](../model_doc/funnel), [GPT-Sw3](../model_doc/gpt-sw3), [OpenAI GPT-2](../model_doc/gpt2), [GPTBigCode](../model_doc/gpt_bigcode), [GPT Neo](../model_doc/gpt_neo), [GPT NeoX](../model_doc/gpt_neox), [I-BERT](../model_doc/ibert), [LayoutLM](../model_doc/layoutlm), [LayoutLMv2](../model_doc/layoutlmv2), [LayoutLMv3](../model_doc/layoutlmv3), [LiLT](../model_doc/lilt), [Longformer](../model_doc/longformer), [LUKE](../model_doc/luke), [MarkupLM](../model_doc/markuplm), [MEGA](../model_doc/mega), [Megatron-BERT](../model_doc/megatron-bert), [MobileBERT](../model_doc/mobilebert), [MPNet](../model_doc/mpnet), [MPT](../model_doc/mpt), [MRA](../model_doc/mra), [Nezha](../model_doc/nezha), [Nyströmformer](../model_doc/nystromformer), [QDQBert](../model_doc/qdqbert), [RemBERT](../model_doc/rembert), [RoBERTa](../model_doc/roberta), [RoBERTa-PreLayerNorm](../model_doc/roberta-prelayernorm), [RoCBert](../model_doc/roc_bert), [RoFormer](../model_doc/roformer), [SqueezeBERT](../model_doc/squeezebert), [XLM](../model_doc/xlm), [XLM-RoBERTa](../model_doc/xlm-roberta), [XLM-RoBERTa-XL](../model_doc/xlm-roberta-xl), [XLNet](../model_doc/xlnet), [X-MOD](../model_doc/xmod), [YOSO](../model_doc/yoso)
-
+[ALBERT](../model_doc/albert), [BERT](../model_doc/bert), [BigBird](../model_doc/big_bird), [BioGpt](../model_doc/biogpt), [BLOOM](../model_doc/bloom), [BROS](../model_doc/bros), [CamemBERT](../model_doc/camembert), [CANINE](../model_doc/canine), [ConvBERT](../model_doc/convbert), [Data2VecText](../model_doc/data2vec-text), [DeBERTa](../model_doc/deberta), [DeBERTa-v2](../model_doc/deberta-v2), [DistilBERT](../model_doc/distilbert), [ELECTRA](../model_doc/electra), [ERNIE](../model_doc/ernie), [ErnieM](../model_doc/ernie_m), [ESM](../model_doc/esm), [Falcon](../model_doc/falcon), [FlauBERT](../model_doc/flaubert), [FNet](../model_doc/fnet), [Funnel Transformer](../model_doc/funnel), [GPT-Sw3](../model_doc/gpt-sw3), [OpenAI GPT-2](../model_doc/gpt2), [GPTBigCode](../model_doc/gpt_bigcode), [GPT Neo](../model_doc/gpt_neo), [GPT NeoX](../model_doc/gpt_neox), [I-BERT](../model_doc/ibert), [LayoutLM](../model_doc/layoutlm), [LayoutLMv2](../model_doc/layoutlmv2), [LayoutLMv3](../model_doc/layoutlmv3), [LiLT](../model_doc/lilt), [Longformer](../model_doc/longformer), [LUKE](../model_doc/luke), [MarkupLM](../model_doc/markuplm), [MEGA](../model_doc/mega), [Megatron-BERT](../model_doc/megatron-bert), [MobileBERT](../model_doc/mobilebert), [MPNet](../model_doc/mpnet), [MPT](../model_doc/mpt), [MRA](../model_doc/mra), [Nezha](../model_doc/nezha), [Nyströmformer](../model_doc/nystromformer), [QDQBert](../model_doc/qdqbert), [RemBERT](../model_doc/rembert), [RoBERTa](../model_doc/roberta), [RoBERTa-PreLayerNorm](../model_doc/roberta-prelayernorm), [RoCBert](../model_doc/roc_bert), [RoFormer](../model_doc/roformer), [SqueezeBERT](../model_doc/squeezebert), [XLM](../model_doc/xlm), [XLM-RoBERTa](../model_doc/xlm-roberta), [XLM-RoBERTa-XL](../model_doc/xlm-roberta-xl), [XLNet](../model_doc/xlnet), [X-MOD](../model_doc/xmod), [YOSO](../model_doc/yoso)
 
 <!--End of the generated tip-->
 

docs/source/en/transformers_agents.md

@@ -14,11 +14,11 @@ rendered properly in your Markdown viewer.
 
 -->
 
-# Transformers Agent
+# Transformers Agents
 
 <Tip warning={true}>
 
-Transformers Agent is an experimental API which is subject to change at any time. Results returned by the agents
+Transformers Agents is an experimental API which is subject to change at any time. Results returned by the agents
 can vary as the APIs or underlying models are prone to change.
 
 </Tip>

docs/source/es/tasks/language_modeling.md

@@ -122,7 +122,7 @@ Así es como puedes crear una función de preprocesamiento para convertir la lis
 ...     return tokenizer([" ".join(x) for x in examples["answers.text"]], truncation=True)
 ```
 
-Usa de 🤗 Datasets la función [`map`](https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map) para aplicar la función de preprocesamiento sobre el dataset en su totalidad. Puedes acelerar la función `map` configurando el argumento `batched=True` para procesar múltiples elementos del dataset a la vez y aumentar la cantidad de procesos con `num_proc`. Elimina las columnas que no necesitas:
+Usa de 🤗 Datasets la función [`map`](https://huggingface.co/docs/datasets/process#map) para aplicar la función de preprocesamiento sobre el dataset en su totalidad. Puedes acelerar la función `map` configurando el argumento `batched=True` para procesar múltiples elementos del dataset a la vez y aumentar la cantidad de procesos con `num_proc`. Elimina las columnas que no necesitas:
 
 ```py
 >>> tokenized_eli5 = eli5.map(

docs/source/fr/index.md

@@ -25,7 +25,7 @@ Apprentissage automatique de pointe pour [PyTorch](https://pytorch.org/), [Tenso
 🗣️ **Audio**: reconnaissance automatique de la parole et classification audio.<br>
 🐙 **Multimodalité**: système de question-réponse avec des tableaux ou images, reconnaissance optique de caractères, extraction d'information depuis des documents scannés et classification de vidéo.
 
-🤗 Transformers prend en charge l'interopérabilité entre PyTorch, TensorFlow et JAX. Cela permet d'utiliser un framework différent à chaque étape de la vie d'un modèle, par example entraîner un modèle en trois lignes de code avec un framework, et le charger pour l'inférence avec un autre. Les modèles peuvent également être exportés dans un format comme ONNX et TorchScript pour être déployés dans des environnements de production.
+🤗 Transformers prend en charge l'interopérabilité entre PyTorch, TensorFlow et JAX. Cela permet d'utiliser un framework différent à chaque étape de la vie d'un modèle, par exemple entraîner un modèle en trois lignes de code avec un framework, et le charger pour l'inférence avec un autre. Les modèles peuvent également être exportés dans un format comme ONNX et TorchScript pour être déployés dans des environnements de production.
 
 Rejoignez la communauté grandissante sur le [Hub](https://huggingface.co/models), le [forum](https://discuss.huggingface.co/) ou [Discord](https://discord.com/invite/JfAtkvEtRb) dès aujourd'hui !
 
@@ -407,4 +407,4 @@ Le tableau ci-dessous représente la prise en charge actuelle dans la bibliothè
 |             YOLOS             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             YOSO              |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 
-<!-- End table-->
+<!-- End table-->

docs/source/fr/quicktour.md

@@ -60,7 +60,7 @@ Le [`pipeline`] est le moyen le plus simple d'utiliser un modèle pré-entraîn
 | Traduction                  | Traduit du texte d'un langage à un autre                                                                      | Texte                | pipeline(task="translation")                  |
 | Classification d'image       | Attribue une catégorie à une image                                                                           | Image                | pipeline(task="image-classification")         |
 | Segmentation d'image           | Attribue une catégorie à chaque pixel d'une image (supporte la segmentation sémantique, panoptique et d'instance) | Image                | pipeline(task="image-segmentation")           |
-| Détection d'objects             | Prédit les délimitations et catégories d'objects dans une image                                                | Image                | pipeline(task="object-detection")             |
+| Détection d'objets             | Prédit les délimitations et catégories d'objets dans une image                                                | Image                | pipeline(task="object-detection")             |
 | Classification d'audio       | Attribue une catégorie à un fichier audio                                                                    | Audio                | pipeline(task="audio-classification")         |
 | Reconnaissance automatique de la parole | Extrait le discours d'un fichier audio en texte                                                                  | Audio                | pipeline(task="automatic-speech-recognition") |
 | Question réponse visuels    | Etant données une image et une question, répond correctement à une question sur l'image                                   | Modalités multiples  | pipeline(task="vqa")                          |
@@ -99,7 +99,7 @@ Le [`pipeline`] peut aussi itérer sur un jeu de données entier pour n'importe
 >>> speech_recognizer = pipeline("automatic-speech-recognition", model="facebook/wav2vec2-base-960h")
 ```
 
-Chargez un jeu de données audio (voir le 🤗 Datasets [Quick Start](https://huggingface.co/docs/datasets/quickstart#audio) pour plus de détails) sur lequel vous souhaitez itérer. Pour cet example, nous chargons le jeu de données [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) :
+Chargez un jeu de données audio (voir le 🤗 Datasets [Quick Start](https://huggingface.co/docs/datasets/quickstart#audio) pour plus de détails) sur lequel vous souhaitez itérer. Pour cet exemple, nous chargeons le jeu de données [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) :
 
 ```py
 >>> from datasets import load_dataset, Audio
@@ -155,7 +155,7 @@ Utilisez [`TFAutoModelForSequenceClassification`] et [`AutoTokenizer`] pour char
 </tf>
 </frameworkcontent>
 
-Specifiez le modèle et le tokenizer dans le [`pipeline`], et utilisez le `classifier` sur le texte en français :
+Spécifiez le modèle et le tokenizer dans le [`pipeline`], et utilisez le `classifier` sur le texte en français :
 
 ```py
 >>> classifier = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)
@@ -418,7 +418,7 @@ En fonction de votre tâche, vous passerez généralement les paramètres suivan
    >>> model = AutoModelForSequenceClassification.from_pretrained("distilbert-base-uncased")
    ```
 
-2. [`TrainingArguments`] contient les hyperparamètres du modèle que vous pouvez changer comme le taux d'apprentissage, la taille due l'échantillon, et le nombre d'époques pour s'entraîner. Les valeurs par défaut sont utilisées si vous ne spécifiez pas d'hyperparamètres d'apprentissage :
+2. [`TrainingArguments`] contient les hyperparamètres du modèle que vous pouvez changer comme le taux d'apprentissage, la taille de l'échantillon, et le nombre d'époques pour s'entraîner. Les valeurs par défaut sont utilisées si vous ne spécifiez pas d'hyperparamètres d'apprentissage :
 
    ```py
    >>> from transformers import TrainingArguments
@@ -547,4 +547,4 @@ Tous les modèles sont des modèles standard [`tf.keras.Model`](https://www.tens
 
 ## Et après ?
 
-Maintenant que vous avez terminé la visite rapide de 🤗 Transformers, consultez nos guides et apprenez à faire des choses plus spécifiques comme créer un modèle personnalisé, finetuner un modèle pour une tâche, et comment entraîner un modèle avec un script. Si vous souhaitez en savoir plus sur les concepts fondamentaux de 🤗 Transformers, jetez un œil à nos guides conceptuels !
+Maintenant que vous avez terminé la visite rapide de 🤗 Transformers, consultez nos guides et apprenez à faire des choses plus spécifiques comme créer un modèle personnalisé, finetuner un modèle pour une tâche, et comment entraîner un modèle avec un script. Si vous souhaitez en savoir plus sur les concepts fondamentaux de 🤗 Transformers, jetez un œil à nos guides conceptuels !

docs/source/ko/_toctree.yml

@@ -25,6 +25,8 @@
     title: 만든 모델 공유하기
   - local: transformers_agents
     title: 에이전트
+  - local: llm_tutorial
+    title: 대규모 언어 모델로 생성하기
   title: 튜토리얼
 - sections:
   - sections:
@@ -147,8 +149,8 @@
       title: TensorFlow 모델을 위한 XLA 통합
   title: (번역중) 성능 및 확장성
 - sections:
-    - local: in_translation
-      title: (번역중) How to contribute to transformers?
+    - local: contributing
+      title: 🤗 Transformers에 기여하는 방법
     - local: add_new_model
       title: 🤗 Transformers에 새로운 모델을 추가하는 방법 
     - local: add_tensorflow_model
@@ -335,8 +337,10 @@
         title: (번역중) Jukebox
       - local: in_translation
         title: (번역중) LED
-      - local: in_translation
-        title: (번역중) LLaMA
+      - local: model_doc/llama
+        title: LLaMA
+      - local: model_doc/llama2
+        title: LLaMA2
       - local: in_translation
         title: (번역중) Longformer
       - local: in_translation

docs/source/ko/contributing.md

@@ -0,0 +1,332 @@
+<!---
+Copyright 2020 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+-->
+
+# 🤗 Transformers에 기여하기 [[contribute-to-transformers]]
+
+누구나 🤗 Transformers에 기여할 수 있으며, 우리는 모든 사람의 기여를 소중히 생각합니다. 코드 기여는 커뮤니티를 돕는 유일한 방법이 아닙니다. 질문에 답하거나 다른 사람을 도와 문서를 개선하는 것도 매우 가치가 있습니다.
+
+🤗 Transformers를 널리 알리는 것도 큰 도움이 됩니다! 멋진 프로젝트들을 가능하게 한 🤗 Transformers 라이브러리에 대해 블로그 게시글에 언급하거나, 도움이 되었을 때마다 Twitter에 알리거나, 저장소에 ⭐️ 를 표시하여 감사 인사를 전해주세요.
+
+어떤 방식으로 기여하든 [행동 규칙](https://github.com/huggingface/transformers/blob/main/CODE_OF_CONDUCT.md)을 숙지하고 존중해주세요.
+
+**이 안내서는 멋진 [scikit-learn 기여 안내서](https://github.com/scikit-learn/scikit-learn/blob/main/CONTRIBUTING.md)에서 큰 영감을 받았습니다.**
+
+## 기여하는 방법 [[ways-to-contribute]]
+
+여러 가지 방법으로 🤗 Transformers에 기여할 수 있습니다:
+
+* 기존 코드의 미해결된 문제를 수정합니다.
+* 버그 또는 새로 추가되길 원하는 기능과 관련된 이슈를 제출합니다.
+* 새로운 모델을 구현합니다.
+* 예제나 문서에 기여합니다.
+
+어디서부터 시작할지 모르겠다면, [Good First Issue](https://github.com/huggingface/transformers/contribute) 목록을 확인해보세요. 이 목록은 초보자도 참여하기 쉬운 오픈 이슈 목록을 제공하며, 당신이 오픈소스에 처음으로 기여하는 데 큰 도움이 될 것입니다. 그저 작업하고 싶은 이슈에 댓글만 달아주면 됩니다.
+
+조금 더 도전적인 작업을 원한다면, [Good Second Issue](https://github.com/huggingface/transformers/labels/Good%20Second%20Issue) 목록도 확인해보세요. 이미 당신이 잘 하고 있다고 생각되더라도, 한 번 시도해보세요! 우리도 여러분을 도울 것입니다. 🚀
+
+> 커뮤니티에 이루어지는 모든 기여는 똑같이 소중합니다. 🥰
+
+## 미해결된 문제 수정하기 [[fixing-outstanding-issues]]
+
+기존 코드에서 발견한 문제점에 대한 해결책이 떠오른 경우, 언제든지 [기여를 시작](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md/#create-a-pull-request)하고 Pull Request를 생성해주세요!
+
+## 버그 관련 이슈를 제기하거나 새로운 기능 요청하기 [[submitting-a-bugrelated-issue-or-feature-request]]
+
+버그 관련 이슈를 제기하거나 새로운 기능을 요청할 때는 다음 가이드라인을 최대한 준수해주세요. 이렇게 하면 좋은 피드백과 함께 빠르게 답변해 드릴 수 있습니다.
+
+### 버그를 발견하셨나요? [[did-you-find-a-bug]]
+
+🤗 Transformers 라이브러리는 사용 중에 겪는 문제를 보고해주는 사용자들 덕분에 더욱 견고해지고 신뢰할 수 있게 되었습니다.
+
+이슈를 보고하기 전에, 버그가 이미 **보고되지 않았는지** 확인해주세요. (GitHub의 이슈 탭 아래의 검색 바를 사용하세요). 이슈는 라이브러리 자체에서 발생한 버그어야 하며, 코드의 다른 부분과 관련된 것이 아니어야 합니다. 버그가 라이브러리의 문제로 발생하였는지 확실하지 않은 경우 먼저 [포럼](https://discuss.huggingface.co/)에서 질문해 주세요. 이렇게 하면 일반적인 질문보다 라이브러리와 관련된 문제를 더 빠르게 해결할 수 있습니다.
+
+버그가 이미 보고되지 않았다는 것을 확인했다면, 다음 정보를 포함하여 이슈를 제출해 주세요. 그러면 우리가 빠르게 해결할 수 있습니다:
+
+* 사용 중인 **운영체제 종류와 버전**, 그리고 **Python**, **PyTorch** 또는 **TensorFlow** 버전.
+* 버그를 30초 이내로 재현할 수 있는 간단하고 독립적인 코드 스니펫.
+* 예외가 발생한 경우 *전체* 트레이스백.
+* 스크린샷과 같이 도움이 될 것으로 생각되는 추가 정보를 첨부해 주세요.
+
+운영체제와 소프트웨어 버전을 자동으로 가져오려면 다음 명령을 실행하세요:
+
+```bash
+transformers-cli env
+```
+
+저장소의 루트 디렉터리에서도 같은 명령을 실행할 수 있습니다:
+
+```bash
+python src/transformers/commands/transformers_cli.py env
+```
+
+
+### 새로운 기능을 원하시나요? [[do-you-want-a-new-feature]]
+
+🤗 Transformers에서 사용하고 싶은 새로운 기능이 있다면, 다음 내용을 포함하여 이슈를 제출해 주세요:
+
+1. 이 기능이 필요한 *이유*는 무엇인가요? 라이브러리에 대한 문제나 불만과 관련이 있나요? 프로젝트에 필요한 기능인가요? 커뮤니티에 도움이 될 만한 기능인가요?
+
+   어떤 내용이든 여러분의 이야기를 듣고 싶습니다!
+
+2. 요청하는 기능을 최대한 자세히 설명해 주세요. 더 많은 정보를 제공할수록 더 나은 도움을 드릴 수 있습니다.
+3. 해당 기능의 사용법을 보여주는 *코드 스니펫*을 제공해 주세요.
+4. 기능과 관련된 논문이 있는 경우 링크를 포함해 주세요.
+
+이슈가 잘 작성되었다면 이슈가 생성된 순간, 이미 80% 정도의 작업이 완료된 것입니다. 
+
+이슈를 제기하는 데 도움이 될 만한 [템플릿](https://github.com/huggingface/transformers/tree/main/templates)도 준비되어 있습니다.
+
+## 새로운 모델을 구현하고 싶으신가요? [[do-you-want-to-implement-a-new-model]]
+
+새로운 모델은 계속해서 출시됩니다. 만약 여러분이 새로운 모델을 구현하고 싶다면 다음 정보를 제공해 주세요.
+
+* 모델에 대한 간단한 설명과 논문 링크.
+* 구현이 공개되어 있다면 구현 링크.
+* 모델 가중치가 사용 가능하다면 가중치 링크.
+
+만약 모델을 직접 기여하고 싶으시다면, 알려주세요. 🤗 Transformers에 추가할 수 있도록 도와드리겠습니다!
+
+새로운 모델을 추가하는 방법에 대한 [상세 안내서와 템플릿](https://github.com/huggingface/transformers/tree/main/templates)을 제공하고 있으며, [🤗 Transformers에 새로운 모델을 추가하는 방법](https://huggingface.co/docs/transformers/add_new_model)에 대한 기술적인 안내서도 있습니다.
+
+## 문서를 추가하고 싶으신가요? [[do-you-want-to-add-documentation]]
+
+우리는 언제나 더 명확하고 정확한 문서를 제공하기 위하여 개선점을 찾고 있습니다. 오탈자나 부족한 내용, 분명하지 않거나 부정확한 내용 등을 알려주시면 개선하는 데 도움이 됩니다. 관심이 있으시다면 변경하거나 기여하실 수 있도록 도와드리겠습니다!
+
+문서를 생성, 빌드 및 작성하는 방법에 대한 자세한 내용은 [README](https://github.com/huggingface/transformers/tree/main/docs) 문서를 확인해 주세요.
+
+## 풀 리퀘스트(Pull Request) 생성하기 [[create-a-pull-request]]
+
+코드를 작성하기 전에 기존의 Pull Request나 이슈를 검색하여 누군가 이미 동일한 작업을 하고 있는지 확인하는 것이 좋습니다. 확실하지 않다면 피드백을 받기 위해 이슈를 열어보는 것이 좋습니다.
+
+🤗 Transformers에 기여하기 위해서는 기본적인 `git` 사용 능력이 필요합니다. `git`은 사용하기 쉬운 도구는 아니지만, 매우 훌륭한 매뉴얼을 제공합니다. 쉘(shell)에서 `git --help`을 입력하여 확인해보세요! 만약 책을 선호한다면, [Pro Git](https://git-scm.com/book/en/v2)은 매우 좋은 참고 자료가 될 것입니다.
+
+🤗 Transformers에 기여하려면 **[Python 3.8]((https://github.com/huggingface/transformers/blob/main/setup.py#L426))** 이상의 버전이 필요합니다. 기여를 시작하려면 다음 단계를 따르세요:
+
+1. 저장소 페이지에서 **[Fork](https://github.com/huggingface/transformers/fork)** 버튼을 클릭하여 저장소를 포크하세요. 이렇게 하면 코드의 복사본이 여러분의 GitHub 사용자 계정 아래에 생성됩니다.
+
+2. 포크한 저장소를 로컬 디스크로 클론하고, 기본 저장소를 원격(remote)으로 추가하세요:
+
+   ```bash
+   git clone git@github.com:<your Github handle>/transformers.git
+   cd transformers
+   git remote add upstream https://github.com/huggingface/transformers.git
+   ```
+
+3. 개발 변경 사항을 저장할 새 브랜치를 생성하세요:
+
+   ```bash
+   git checkout -b a-descriptive-name-for-my-changes
+   ```
+
+   🚨 절대 `main` 브랜치에서 작업하지 **마세요!**
+
+4. 가상 환경에서 다음 명령을 실행하여 개발 환경을 설정하세요:
+
+   ```bash
+   pip install -e ".[dev]"
+   ```
+
+   만약 이미 가상 환경에 🤗 Transformers가 설치되어 있다면, `-e` 플래그를 사용하여 설치하기 전에 `pip uninstall transformers`로 제거해주세요.
+   
+   여러분의 운영체제에 따라서, 그리고 🤗 Transformers의 선택적 의존성의 수가 증가하면서, 이 명령이 실패할 수도 있습니다. 그럴 경우 사용하려는 딥러닝 프레임워크(PyTorch, TensorFlow, 그리고/또는 Flax)를 설치한 후 아래 명령을 실행해주세요:
+
+   ```bash
+   pip install -e ".[quality]"
+   ```
+
+   대부분의 경우 이것으로 충분할 것입니다.
+
+5. 브랜치에서 기능을 개발하세요.
+
+   코드를 작업하는 동안 테스트 스위트(test suite)가 통과하는지 확인하세요. 다음과 같이 변경 사항에 영향을 받는 테스트를 실행하세요:
+
+   ```bash
+   pytest tests/<TEST_TO_RUN>.py
+   ```
+
+   테스트에 대한 더 많은 정보는 [테스트](https://huggingface.co/docs/transformers/testing) 가이드를 확인하세요.
+
+   🤗 Transformers는 `black`과 `ruff`를 사용하여 소스 코드의 형식을 일관되게 유지합니다. 변경 사항을 적용한 후에는 다음 명령으로 자동으로 스타일 교정 및 코드 검증을 수행하세요:
+
+   ```bash
+   make fixup
+   ```
+
+   이것은 또한 작업 중인 PR에서 수정한 파일에서만 작동하도록 최적화되어 있습니다.
+
+   검사를 하나씩 실행하려는 경우, 다음 명령으로 스타일 교정을 적용할 수 있습니다:
+
+   ```bash
+   make style
+   ```
+
+   🤗 Transformers는 또한 `ruff`와 몇 가지 사용자 정의 스크립트를 사용하여 코딩 실수를 확인합니다. CI를 통해 품질 관리가 수행되지만, 다음 명령으로 동일한 검사를 실행할 수 있습니다:
+
+   ```bash
+   make quality
+   ```
+
+   마지막으로, 새 모델을 추가할 때 일부 파일을 업데이트하는 것을 잊지 않도록 하기 위한 많은 스크립트가 있습니다. 다음 명령으로 이러한 스크립트를 실행할 수 있습니다:
+
+   ```bash
+   make repo-consistency
+   ```
+
+   이러한 검사에 대해 자세히 알아보고 관련 문제를 해결하는 방법은 [Pull Request에 대한 검사](https://huggingface.co/docs/transformers/pr_checks) 가이드를 확인하세요.
+
+   만약 `docs/source` 디렉터리 아래의 문서를 수정하는 경우, 문서가 빌드될 수 있는지 확인하세요. 이 검사는 Pull Request를 열 때도 CI에서 실행됩니다. 로컬 검사를 실행하려면 문서 빌더를 설치해야 합니다:
+   
+   ```bash
+   pip install ".[docs]"
+   ```
+
+   저장소의 루트 디렉터리에서 다음 명령을 실행하세요:
+
+   ```bash
+   doc-builder build transformers docs/source/en --build_dir ~/tmp/test-build
+   ```
+
+   이 명령은 `~/tmp/test-build` 폴더에 문서를 빌드하며, 생성된 Markdown 파일을 선호하는 편집기로 확인할 수 있습니다. Pull Request를 열 때 GitHub에서 문서를 미리 볼 수도 있습니다.
+
+   변경 사항에 만족하면 `git add`로 변경된 파일을 추가하고, `git commit`으로 변경 사항을 로컬에 기록하세요:
+
+   ```bash
+   git add modified_file.py
+   git commit
+   ```
+
+   [좋은 커밋 메시지](https://chris.beams.io/posts/git-commit/)를 작성하여 변경 사항을 명확하게 전달하세요!
+
+   변경 사항을 프로젝트 원본 저장소와 동기화하려면, PR을 *열기 전에* 브랜치를 `upstream/branch`로 리베이스(rebase)하세요. 또는 관리자의 요청에 이 작업이 필요할 수 있습니다:
+
+   ```bash
+   git fetch upstream
+   git rebase upstream/main
+   ```
+   
+   변경 사항을 브랜치에 푸시하세요:
+
+   ```bash
+   git push -u origin a-descriptive-name-for-my-changes
+   ```
+
+   이미 PR을 열었다면, `--force` 플래그와 함께 강제 푸시해야 합니다. 아직 PR이 열리지 않았다면 정상적으로 변경 사항을 푸시하면 됩니다.
+
+6. 이제 GitHub에서 포크한 저장소로 이동하고 **Pull request(풀 리퀘스트)**를 클릭하여 Pull Request를 열 수 있습니다. 아래의 [체크리스트](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md/#pull-request-checklist)에서 모든 항목에 체크 표시를 하세요. 준비가 완료되면 프로젝트 관리자에게 변경 사항을 보내 검토를 요청할 수 있습니다.
+
+7. 관리자가 변경 사항을 요청해도 괜찮습니다. 핵심 기여자들도 동일한 상황을 겪습니다! 모두가 변경 사항을 Pull Request에서 볼 수 있도록, 로컬 브랜치에서 작업하고 변경 사항을 포크한 저장소로 푸시하세요. 그러면 변경 사항이 자동으로 Pull Request에 나타납니다.
+
+### Pull Request 체크리스트 [[pull-request-checklist]]
+
+☐ Pull Request 제목은 기여 내용을 요약해야 합니다.<br>
+☐ Pull Request가 이슈를 해결하는 경우, Pull Request 설명에 이슈 번호를 언급하여 연관되어 있음을 알려주세요. (이슈를 확인하는 사람들이 해당 이슈에 대한 작업이 진행 중임을 알 수 있게 합니다).<br>
+☐ 작업이 진행중이라면 제목 앞에 `[WIP]`를 붙여주세요. 중복 작업을 피하고 병합할 준비가 된 PR과 구분하기에 유용합니다.<br>
+☐ 기존 테스트를 통과하는지 확인하세요.<br>
+☐ 새로운 기능을 추가하는 경우, 해당 기능에 대한 테스트도 추가하세요.<br>
+   - 새 모델을 추가하는 경우, `ModelTester.all_model_classes = (MyModel, MyModelWithLMHead,...)`을 사용하여 일반적인 테스트를 활성화하세요.
+   - 새 `@slow` 테스트를 추가하는 경우, 다음 명령으로 테스트를 통과하는지 확인하세요: `RUN_SLOW=1 python -m pytest tests/models/my_new_model/test_my_new_model.py`.
+   - 새 토크나이저를 추가하는 경우, 테스트를 작성하고 다음 명령으로 테스트를 통과하는지 확인하세요: `RUN_SLOW=1 python -m pytest tests/models/{your_model_name}/test_tokenization_{your_model_name}.py`. 
+   - CircleCI에서는 느린 테스트를 실행하지 않지만, GitHub Actions에서는 매일 밤 실행됩니다!<br>
+
+☐ 모든 공개 메소드는 유용한 기술문서를 가져야 합니다 (예를 들어 [`modeling_bert.py`](https://github.com/huggingface/transformers/blob/main/src/transformers/models/bert/modeling_bert.py) 참조).<br>
+☐ 저장소가 빠르게 성장하고 있으므로 저장소에 상당한 부담을 주는 이미지, 동영상 및 기타 텍스트가 아닌 파일은 추가하지 마세요. 대신 [`hf-internal-testing`](https://huggingface.co/hf-internal-testing)과 같은 Hub 저장소를 사용하여 이러한 파일을 호스팅하고 URL로 참조하세요. 문서와 관련된 이미지는 다음 저장소에 배치하는 것을 권장합니다: [huggingface/documentation-images](https://huggingface.co/datasets/huggingface/documentation-images). 이 데이터셋 저장소에서 PR을 열어서 Hugging Face 멤버에게 병합을 요청할 수 있습니다.
+
+Pull Request에서 실행되는 검사에 대한 자세한 정보는 [Pull Request에 대한 검사](https://huggingface.co/docs/transformers/pr_checks) 가이드를 확인하세요.
+
+### 테스트 [[tests]]
+
+라이브러리 동작과 여러 예제를 테스트할 수 있는 광범위한 테스트 스위트가 포함되어 있습니다. 라이브러리 테스트는 [tests](https://github.com/huggingface/transformers/tree/main/tests) 폴더에, 예제 테스트는 [examples](https://github.com/huggingface/transformers/tree/main/examples) 폴더에 있습니다.
+
+속도가 빠른 `pytest`와 `pytest-xdist`를 선호합니다. 저장소의 루트 디렉터리에서 테스트를 실행할 *하위 폴더 경로 또는 테스트 파일 경로*를 지정하세요.
+
+```bash
+python -m pytest -n auto --dist=loadfile -s -v ./tests/models/my_new_model
+```
+
+마찬가지로 `examples` 디렉터리에서도 *하위 폴더 경로 또는 테스트 파일 경로*를 지정하세요. 예를 들어, 다음 명령은 PyTorch `examples` 디렉터리의 텍스트 분류 하위 폴더를 테스트합니다:
+
+```bash
+pip install -r examples/xxx/requirements.txt  # only needed the first time
+python -m pytest -n auto --dist=loadfile -s -v ./examples/pytorch/text-classification
+```
+
+이것이 실제로 `make test` 및 `make test-examples` 명령이 구현되는 방식입니다 (`pip install`은 제외합니다)!
+
+또한 특정 기능만 테스트하기 위한 더 작은 테스트를 지정할 수 있습니다.
+
+기본적으로 느린 테스트는 건너뛰지만 `RUN_SLOW` 환경 변수를 `yes`로 설정하여 실행할 수 있습니다. 이렇게 하면 많은 기가바이트 단위의 모델이 다운로드되므로 충분한 디스크 공간, 좋은 인터넷 연결과 많은 인내가 필요합니다!
+
+<Tip warning={true}>
+
+테스트를 실행하려면 *하위 폴더 경로 또는 테스트 파일 경로*를 지정하세요. 그렇지 않으면 `tests` 또는 `examples` 폴더의 모든 테스트를 실행하게 되어 매우 긴 시간이 걸립니다!
+
+</Tip>
+
+```bash
+RUN_SLOW=yes python -m pytest -n auto --dist=loadfile -s -v ./tests/models/my_new_model
+RUN_SLOW=yes python -m pytest -n auto --dist=loadfile -s -v ./examples/pytorch/text-classification
+```
+
+느린 테스트와 마찬가지로, 다음과 같이 테스트 중에 기본적으로 활성화되지 않는 다른 환경 변수도 있습니다:
+- `RUN_CUSTOM_TOKENIZERS`: 사용자 정의 토크나이저 테스트를 활성화합니다.
+- `RUN_PT_FLAX_CROSS_TESTS`: PyTorch + Flax 통합 테스트를 활성화합니다.
+- `RUN_PT_TF_CROSS_TESTS`: TensorFlow + PyTorch 통합 테스트를 활성화합니다.
+
+더 많은 환경 변수와 추가 정보는 [testing_utils.py](src/transformers/testing_utils.py)에서 찾을 수 있습니다.
+
+🤗 Transformers는 테스트 실행기로 `pytest`를 사용합니다. 그러나 테스트 스위트 자체에서는 `pytest` 관련 기능을 사용하지 않습니다.
+
+이것은 `unittest`가 완전히 지원된다는 것을 의미합니다. 다음은 `unittest`로 테스트를 실행하는 방법입니다:
+
+```bash
+python -m unittest discover -s tests -t . -v
+python -m unittest discover -s examples -t examples -v
+```
+
+### 스타일 가이드 [[style-guide]]
+
+문서는 [Google Python 스타일 가이드](https://google.github.io/styleguide/pyguide.html)를 따릅니다. 자세한 정보는 [문서 작성 가이드](https://github.com/huggingface/transformers/tree/main/docs#writing-documentation---specification)를 확인하세요.
+
+### Windows에서 개발 [[develop-on-windows]]
+
+Windows에서 개발할 경우([Windows Subsystem for Linux](https://learn.microsoft.com/en-us/windows/wsl/) 또는 WSL에서 작업하지 않는 한) Windows `CRLF` 줄 바꿈을 Linux `LF` 줄 바꿈으로 변환하도록 git을 구성해야 합니다:
+
+```bash
+git config core.autocrlf input
+```
+
+Windows에서 `make` 명령을 실행하는 한 가지 방법은 MSYS2를 사용하는 것입니다:
+
+1. [MSYS2](https://www.msys2.org/)를 다운로드합니다. `C:\msys64`에 설치되었다고 가정합니다.
+2. CLI에서 `C:\msys64\msys2.exe`를 엽니다 (시작 메뉴에서 사용 가능해야 함).
+3. 쉘에서 다음을 실행하여: `pacman -Syu` 및 `pacman -S make`로 `make`를 설치합니다.
+4. 환경 변수 PATH에 `C:\msys64\usr\bin`을 추가하세요.
+
+이제 모든 터미널 (Powershell, cmd.exe 등)에서 `make`를 사용할 수 있습니다! 🎉
+
+### 포크한 저장소를 상위 원본 브랜치(main)과 동기화하기 (Hugging Face 저장소) [[sync-a-forked-repository-with-upstream-main-the-hugging-face-repository]]
+
+포크한 저장소의 main 브랜치를 업데이트할 때, 다음 단계를 따라 수행해주세요. 이렇게 하면 각 upstream PR에 참조 노트가 추가되는 것을 피하고 이러한 PR에 관여하는 개발자들에게 불필요한 알림이 전송되는 것을 방지할 수 있습니다.
+
+1. 가능하면 포크된 저장소의 브랜치 및 PR을 사용하여 upstream과 동기화하지 마세요. 대신 포크된 main 저장소에 직접 병합하세요.
+2. PR이 반드시 필요한 경우, 브랜치를 확인한 후 다음 단계를 사용하세요:
+
+```bash
+git checkout -b your-branch-for-syncing
+git pull --squash --no-commit upstream main
+git commit -m '<your message without GitHub references>'
+git push --set-upstream origin your-branch-for-syncing
+```

docs/source/ko/custom_tools.md

@@ -20,7 +20,7 @@ Transformers와 관련하여 어떤 도구와 에이전트가 있는지 잘 모
 
 <Tip warning={true}>
 
-Transformers Agent는 실험 중인 API로 언제든지 변경될 수 있습니다. 
+Transformers Agents는 실험 중인 API로 언제든지 변경될 수 있습니다. 
 API 또는 기반 모델이 변경되기 쉽기 때문에 에이전트가 반환하는 결과도 달라질 수 있습니다.
 
 </Tip>

docs/source/ko/llm_tutorial.md

@@ -0,0 +1,221 @@
+<!--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.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+
+# 대규모 언어 모델로 생성하기 [[generation-with-llms]]
+
+[[open-in-colab]]
+
+LLM 또는 대규모 언어 모델은 텍스트 생성의 핵심 구성 요소입니다. 간단히 말하면, 주어진 입력 텍스트에 대한 다음 단어(정확하게는 토큰)를 예측하기 위해 훈련된 대규모 사전 훈련 변환기 모델로 구성됩니다. 토큰을 한 번에 하나씩 예측하기 때문에 새로운 문장을 생성하려면 모델을 호출하는 것 외에 더 복잡한 작업을 수행해야 합니다. 즉, 자기회귀 생성을 수행해야 합니다.
+
+자기회귀 생성은 몇 개의 초기 입력값을 제공한 후, 그 출력을 다시 모델에 입력으로 사용하여 반복적으로 호출하는 추론 과정입니다. 🤗 Transformers에서는 [`~generation.GenerationMixin.generate`] 메소드가 이 역할을 하며, 이는 생성 기능을 가진 모든 모델에서 사용 가능합니다.
+
+이 튜토리얼에서는 다음 내용을 다루게 됩니다:
+
+* LLM으로 텍스트 생성
+* 일반적으로 발생하는 문제 해결
+* LLM을 최대한 활용하기 위한 다음 단계
+
+시작하기 전에 필요한 모든 라이브러리가 설치되어 있는지 확인하세요:
+
+```bash
+pip install transformers bitsandbytes>=0.39.0 -q
+```
+
+
+## 텍스트 생성 [[generate-text]]
+
+[인과적 언어 모델링(causal language modeling)](tasks/language_modeling)을 목적으로 학습된 언어 모델은 일련의 텍스트 토큰을 입력으로 사용하고, 그 결과로 다음 토큰이 나올 확률 분포를 제공합니다.
+
+<!-- [GIF 1 -- FWD PASS] -->
+<figure class="image table text-center m-0 w-full">
+    <video
+        style="max-width: 90%; margin: auto;"
+        autoplay loop muted playsinline
+        src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/assisted-generation/gif_1_1080p.mov"
+    ></video>
+    <figcaption>"LLM의 전방 패스"</figcaption>
+</figure>
+
+LLM과 자기회귀 생성을 함께 사용할 때 핵심적인 부분은 이 확률 분포로부터 다음 토큰을 어떻게 고를 것인지입니다. 다음 반복 과정에 사용될 토큰을 결정하는 한, 어떠한 방법도 가능합니다. 확률 분포에서 가장 가능성이 높은 토큰을 선택하는 것처럼 간단할 수도 있고, 결과 분포에서 샘플링하기 전에 수십 가지 변환을 적용하는 것처럼 복잡할 수도 있습니다.
+
+<!-- [GIF 2 -- TEXT GENERATION] -->
+<figure class="image table text-center m-0 w-full">
+    <video
+        style="max-width: 90%; margin: auto;"
+        autoplay loop muted playsinline
+        src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/assisted-generation/gif_2_1080p.mov"
+    ></video>
+    <figcaption>"자기회귀 생성은 확률 분포에서 다음 토큰을 반복적으로 선택하여 텍스트를 생성합니다."</figcaption>
+</figure>
+
+위에서 설명한 과정은 어떤 종료 조건이 충족될 때까지 반복적으로 수행됩니다. 모델이 시퀀스의 끝(EOS 토큰)을 출력할 때까지를 종료 조건으로 하는 것이 이상적입니다. 그렇지 않은 경우에는 미리 정의된 최대 길이에 도달했을 때 생성이 중단됩니다.
+
+모델이 예상대로 동작하기 위해선 토큰 선택 단계와 정지 조건을 올바르게 설정하는 것이 중요합니다. 이러한 이유로, 각 모델에는 기본 생성 설정이 잘 정의된 [`~generation.GenerationConfig`] 파일이 함께 제공됩니다.
+
+코드를 확인해봅시다!
+
+<Tip>
+
+기본 LLM 사용에 관심이 있다면, 우리의 [`Pipeline`](pipeline_tutorial) 인터페이스로 시작하는 것을 추천합니다. 그러나 LLM은 양자화나 토큰 선택 단계에서의 미세한 제어와 같은 고급 기능들을 종종 필요로 합니다. 이러한 작업은 [`~generation.GenerationMixin.generate`]를 통해 가장 잘 수행될 수 있습니다. LLM을 이용한 자기회귀 생성은 자원을 많이 소모하므로, 적절한 처리량을 위해 GPU에서 실행되어야 합니다.
+
+</Tip>
+
+<!-- TODO: update example to llama 2 (or a newer popular baseline) when it becomes ungated -->
+먼저, 모델을 불러오세요.
+
+```py
+>>> from transformers import AutoModelForCausalLM
+
+>>> model = AutoModelForCausalLM.from_pretrained(
+...     "openlm-research/open_llama_7b", device_map="auto", load_in_4bit=True
+... )
+```
+
+`from_pretrained` 함수를 호출할 때 2개의 플래그를 주목하세요:
+
+- `device_map`은 모델이 GPU로 이동되도록 합니다.
+- `load_in_4bit`는 리소스 요구 사항을 크게 줄이기 위해 [4비트 동적 양자화](main_classes/quantization)를 적용합니다.
+
+이 외에도 모델을 초기화하는 다양한 방법이 있지만, LLM을 처음 시작할 때 이 설정을 추천합니다.
+
+이어서 텍스트 입력을 [토크나이저](tokenizer_summary)으로 전처리하세요.
+
+```py
+>>> from transformers import AutoTokenizer
+
+>>> tokenizer = AutoTokenizer.from_pretrained("openlm-research/open_llama_7b")
+>>> model_inputs = tokenizer(["A list of colors: red, blue"], return_tensors="pt").to("cuda")
+```
+
+`model_inputs` 변수에는 토큰화된 텍스트 입력과 함께 어텐션 마스크가 들어 있습니다. [`~generation.GenerationMixin.generate`]는 어텐션 마스크가 제공되지 않았을 경우에도 이를 추론하려고 노력하지만, 최상의 성능을 위해서는 가능하면 어텐션 마스크를 전달하는 것을 권장합니다. 
+
+마지막으로 [`~generation.GenerationMixin.generate`] 메소드를 호출해 생성된 토큰을 얻은 후, 이를 출력하기 전에 텍스트 형태로 변환하세요.
+
+```py
+>>> generated_ids = model.generate(**model_inputs)
+>>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+'A list of colors: red, blue, green, yellow, black, white, and brown'
+```
+
+이게 전부입니다! 몇 줄의 코드만으로 LLM의 능력을 활용할 수 있게 되었습니다.
+
+
+## 일반적으로 발생하는 문제 [[common-pitfalls]]
+
+[생성 전략](generation_strategies)이 많고, 기본값이 항상 사용 사례에 적합하지 않을 수 있습니다. 출력이 예상과 다를 때 흔히 발생하는 문제와 이를 해결하는 방법에 대한 목록을 만들었습니다.
+
+```py
+>>> from transformers import AutoModelForCausalLM, AutoTokenizer
+
+>>> tokenizer = AutoTokenizer.from_pretrained("openlm-research/open_llama_7b")
+>>> tokenizer.pad_token = tokenizer.eos_token  # Llama has no pad token by default
+>>> model = AutoModelForCausalLM.from_pretrained(
+...     "openlm-research/open_llama_7b", device_map="auto", load_in_4bit=True
+... )
+```
+
+### 생성된 출력이 너무 짧거나 길다 [[generated-output-is-too-shortlong]]
+
+[`~generation.GenerationConfig`] 파일에서 별도로 지정하지 않으면, `generate`는 기본적으로 최대 20개의 토큰을 반환합니다. `generate` 호출에서 `max_new_tokens`을 수동으로 설정하여 반환할 수 있는 새 토큰의 최대 수를 설정하는 것이 좋습니다. LLM(정확하게는 [디코더 전용 모델](https://huggingface.co/learn/nlp-course/chapter1/6?fw=pt))은 입력 프롬프트도 출력의 일부로 반환합니다.
+
+
+```py
+>>> model_inputs = tokenizer(["A sequence of numbers: 1, 2"], return_tensors="pt").to("cuda")
+
+>>> # By default, the output will contain up to 20 tokens
+>>> generated_ids = model.generate(**model_inputs)
+>>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+'A sequence of numbers: 1, 2, 3, 4, 5'
+
+>>> # Setting `max_new_tokens` allows you to control the maximum length
+>>> generated_ids = model.generate(**model_inputs, max_new_tokens=50)
+>>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+'A sequence of numbers: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16,'
+```
+
+### 잘못된 생성 모드 [[incorrect-generation-mode]]
+
+기본적으로 [`~generation.GenerationConfig`] 파일에서 별도로 지정하지 않으면, `generate`는 각 반복에서 가장 확률이 높은 토큰을 선택합니다(그리디 디코딩). 하려는 작업에 따라 이 방법은 바람직하지 않을 수 있습니다. 예를 들어, 챗봇이나 에세이 작성과 같은 창의적인 작업은 샘플링이 적합할 수 있습니다. 반면, 오디오를 텍스트로 변환하거나 번역과 같은 입력 기반 작업은 그리디 디코딩이 더 적합할 수 있습니다. `do_sample=True`로 샘플링을 활성화할 수 있으며, 이 주제에 대한 자세한 내용은 이 [블로그 포스트](https://huggingface.co/blog/how-to-generate)에서 볼 수 있습니다.
+
+```py
+>>> # Set seed or reproducibility -- you don't need this unless you want full reproducibility
+>>> from transformers import set_seed
+>>> set_seed(0)
+
+>>> model_inputs = tokenizer(["I am a cat."], return_tensors="pt").to("cuda")
+
+>>> # LLM + greedy decoding = repetitive, boring output
+>>> generated_ids = model.generate(**model_inputs)
+>>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+'I am a cat. I am a cat. I am a cat. I am a cat'
+
+>>> # With sampling, the output becomes more creative!
+>>> generated_ids = model.generate(**model_inputs, do_sample=True)
+>>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+'I am a cat.\nI just need to be. I am always.\nEvery time'
+```
+
+### 잘못된 패딩 [[wrong-padding-side]]
+
+LLM은 [디코더 전용](https://huggingface.co/learn/nlp-course/chapter1/6?fw=pt) 구조를 가지고 있어, 입력 프롬프트에 대해 지속적으로 반복 처리를 합니다. 입력 데이터의 길이가 다르면 패딩 작업이 필요합니다. LLM은 패딩 토큰에서 작동을 이어가도록 설계되지 않았기 때문에, 입력 왼쪽에 패딩이 추가 되어야 합니다. 그리고 어텐션 마스크도 꼭 `generate` 함수에 전달되어야 합니다!
+
+```py
+>>> # The tokenizer initialized above has right-padding active by default: the 1st sequence,
+>>> # which is shorter, has padding on the right side. Generation fails.
+>>> model_inputs = tokenizer(
+...     ["1, 2, 3", "A, B, C, D, E"], padding=True, return_tensors="pt"
+... ).to("cuda")
+>>> generated_ids = model.generate(**model_inputs)
+>>> tokenizer.batch_decode(generated_ids[0], skip_special_tokens=True)[0]
+''
+
+>>> # With left-padding, it works as expected!
+>>> tokenizer = AutoTokenizer.from_pretrained("openlm-research/open_llama_7b", padding_side="left")
+>>> tokenizer.pad_token = tokenizer.eos_token  # Llama has no pad token by default
+>>> model_inputs = tokenizer(
+...     ["1, 2, 3", "A, B, C, D, E"], padding=True, return_tensors="pt"
+... ).to("cuda")
+>>> generated_ids = model.generate(**model_inputs)
+>>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+'1, 2, 3, 4, 5, 6,'
+```
+
+<!-- TODO: when the prompting guide is ready, mention the importance of setting the right prompt in this section -->
+
+## 추가 자료 [[further-resources]]
+
+자기회귀 생성 프로세스는 상대적으로 단순한 편이지만, LLM을 최대한 활용하려면 여러 가지 요소를 고려해야 하므로 쉽지 않을 수 있습니다. LLM에 대한 더 깊은 이해와 활용을 위한 다음 단계는 아래와 같습니다:
+
+<!-- TODO: complete with new guides -->
+### 고급 생성 사용 [[advanced-generate-usage]]
+
+1. [가이드](generation_strategies)는 다양한 생성 방법을 제어하는 방법, 생성 설정 파일을 설정하는 방법, 출력을 스트리밍하는 방법에 대해 설명합니다.
+2. [`~generation.GenerationConfig`]와 [`~generation.GenerationMixin.generate`], [generate-related classes](internal/generation_utils)를 참조해보세요.
+
+### LLM 리더보드 [[llm-leaderboards]]
+
+1. [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)는 오픈 소스 모델의 품질에 중점을 둡니다.
+2. [Open LLM-Perf Leaderboard](https://huggingface.co/spaces/optimum/llm-perf-leaderboard)는 LLM 처리량에 중점을 둡니다.
+
+### 지연 시간 및 처리량 [[latency-and-throughput]]
+
+1. 메모리 요구 사항을 줄이려면, 동적 양자화에 대한 [가이드](main_classes/quantization)를 참조하세요.
+
+### 관련 라이브러리 [[related-libraries]]
+
+1. [`text-generation-inference`](https://github.com/huggingface/text-generation-inference)는 LLM을 위한 실제 운영 환경에 적합한 서버입니다.
+2. [`optimum`](https://github.com/huggingface/optimum)은 특정 하드웨어 장치에서 LLM을 최적화하기 위해 🤗 Transformers를 확장한 것입니다.

docs/source/ko/model_doc/llama.md

@@ -0,0 +1,117 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+# LLaMA [[llama]]
+
+## 개요 [[overview]]
+
+LLaMA 모델은 Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, Guillaume Lample에 의해 제안된 [LLaMA: Open and Efficient Foundation Language Models](https://arxiv.org/abs/2302.13971)에서 소개되었습니다. 이 모델은 7B에서 65B개의 파라미터까지 다양한 크기의 기초 언어 모델을 모아놓은 것입니다.
+
+논문의 초록은 다음과 같습니다:
+
+*"LLaMA는 7B에서 65B개의 파라미터 수를 가진 기초 언어 모델의 모음입니다. 우리는 수조 개의 토큰으로 모델을 훈련시켰고, 공개적으로 이용 가능한 데이터셋만을 사용하여 최고 수준의 모델을 훈련시킬 수 있음을 보여줍니다. 특히, LLaMA-13B 모델은 대부분의 벤치마크에서 GPT-3 (175B)를 능가하며, LLaMA-65B는 최고 수준의 모델인 Chinchilla-70B와 PaLM-540B에 버금가는 성능을 보입니다. 우리는 모든 모델을 연구 커뮤니티에 공개합니다."*
+
+팁:
+
+- LLaMA 모델의 가중치는 [이 양식](https://docs.google.com/forms/d/e/1FAIpQLSfqNECQnMkycAp2jP4Z9TFX0cGR4uf7b_fBxjY_OjhJILlKGA/viewform?usp=send_form)을 작성하여 얻을 수 있습니다.
+- 가중치를 다운로드한 후에는 이를 [변환 스크립트](https://github.com/huggingface/transformers/blob/main/src/transformers/models/llama/convert_llama_weights_to_hf.py)를 사용하여 Hugging Face Transformers 형식으로 변환해야합니다. 변환 스크립트를 실행하려면 아래의 예시 명령어를 참고하세요:
+
+```bash
+python src/transformers/models/llama/convert_llama_weights_to_hf.py \
+    --input_dir /path/to/downloaded/llama/weights --model_size 7B --output_dir /output/path
+```
+
+- 변환을 하였다면 모델과 토크나이저는 다음과 같이 로드할 수 있습니다:
+
+```python
+from transformers import LlamaForCausalLM, LlamaTokenizer
+
+tokenizer = LlamaTokenizer.from_pretrained("/output/path")
+model = LlamaForCausalLM.from_pretrained("/output/path")
+```
+
+스크립트를 실행하기 위해서는 모델을 float16 정밀도로 전부 로드할 수 있을 만큼의 충분한 CPU RAM이 필요합니다. (가장 큰 버전의 모델이 여러 체크포인트로 나뉘어 있더라도, 각 체크포인트는 모델의 각 가중치의 일부를 포함하고 있기 때문에 모든 체크포인트를 RAM에 로드해야 합니다) 65B 모델의 경우, 총 130GB의 RAM이 필요합니다.
+
+
+- LLaMA 토크나이저는 [sentencepiece](https://github.com/google/sentencepiece)를 기반으로 하는 BPE 모델입니다. sentencepiece의 특징 중 하나는 시퀀스를 디코딩할 때 첫 토큰이 단어의 시작이라면 (예를 들어 "Banana"), 토크나이저는 문자열 앞에 공백을 추가하지 않는다는 것입니다.
+
+이 모델은 [BlackSamorez](https://huggingface.co/BlackSamorez)의 기여와 함께, [zphang](https://huggingface.co/zphang)에 의해 제공되었습니다. Hugging Face에서의 구현 코드는 GPT-NeoX를 기반으로 하며 [여기](https://github.com/EleutherAI/gpt-neox)에서 찾을 수 있고, 저자의 코드 원본은 [여기](https://github.com/facebookresearch/llama)에서 확인할 수 있습니다.
+
+
+원래 LLaMA 모델을 기반으로 Meta AI에서 몇 가지 후속 작업을 발표했습니다:
+
+- **Llama2**: Llama2는 구조적인 몇 가지 수정(Grouped Query Attention)을 통해 개선된 버전이며, 2조 개의 토큰으로 사전 훈련이 되어 있습니다. Llama2에 대한 자세한 내용은 [이 문서](llama2)를 참고하세요.
+
+## 리소스 [[resources]]
+
+LLaMA를 시작하는 데 도움이 될 Hugging Face 및 커뮤니티(🌎로 표시)의 공식 자료 목록입니다. 여기에 자료를 제출하고 싶다면 Pull Request를 올려주세요! 추가할 자료는 기존의 자료와 중복되지 않고 새로운 내용을 보여주는 것이 좋습니다.
+
+<PipelineTag pipeline="text-classification"/>
+
+- LLaMA 모델을 텍스트 분류 작업에 적용하기 위한 프롬프트 튜닝 방법에 대한 [노트북](https://colab.research.google.com/github/bigscience-workshop/petals/blob/main/examples/prompt-tuning-sst2.ipynb#scrollTo=f04ba4d2) 🌎
+
+<PipelineTag pipeline="question-answering"/>
+
+- [Stack Exchange](https://stackexchange.com/)에서 질문에 답하는 LLaMA를 훈련하는 방법을 위한 [StackLLaMA: RLHF로 LLaMA를 훈련하는 실전 가이드](https://huggingface.co/blog/stackllama#stackllama-a-hands-on-guide-to-train-llama-with-rlhf) 🌎
+
+⚗️ 최적화
+- 제한된 메모리를 가진 GPU에서 xturing 라이브러리를 사용하여 LLaMA 모델을 미세 조정하는 방법에 대한 [노트북](https://colab.research.google.com/drive/1SQUXq1AMZPSLD4mk3A3swUIc6Y2dclme?usp=sharing) 🌎
+
+⚡️ 추론
+- 🤗 PEFT 라이브러리의 PeftModel을 사용하여 LLaMA 모델을 실행하는 방법에 대한 [노트북](https://colab.research.google.com/github/DominguesM/alpaca-lora-ptbr-7b/blob/main/notebooks/02%20-%20Evaluate.ipynb) 🌎
+- LangChain을 사용하여 PEFT 어댑터 LLaMA 모델을 로드하는 방법에 대한 [노트북](https://colab.research.google.com/drive/1l2GiSSPbajVyp2Nk3CFT4t3uH6-5TiBe?usp=sharing) 🌎
+
+🚀 배포
+- 🤗 PEFT 라이브러리와 사용자 친화적인 UI로 LLaMA 모델을 미세 조정하는 방법에 대한 [노트북](https://colab.research.google.com/github/lxe/simple-llama-finetuner/blob/master/Simple_LLaMA_FineTuner.ipynb#scrollTo=3PM_DilAZD8T) 🌎
+- Amazon SageMaker에서 텍스트 생성을 위해 Open-LLaMA 모델을 배포하는 방법에 대한 [노트북](https://github.com/aws/amazon-sagemaker-examples/blob/main/introduction_to_amazon_algorithms/jumpstart-foundation-models/text-generation-open-llama.ipynb) 🌎
+
+## LlamaConfig [[llamaconfig]]
+
+[[autodoc]] LlamaConfig
+
+
+## LlamaTokenizer [[llamatokenizer]]
+
+[[autodoc]] LlamaTokenizer
+    - build_inputs_with_special_tokens
+    - get_special_tokens_mask
+    - create_token_type_ids_from_sequences
+    - save_vocabulary
+
+## LlamaTokenizerFast [[llamatokenizerfast]]
+
+[[autodoc]] LlamaTokenizerFast
+    - build_inputs_with_special_tokens
+    - get_special_tokens_mask
+    - create_token_type_ids_from_sequences
+    - update_post_processor
+    - save_vocabulary
+
+## LlamaModel [[llamamodel]]
+
+[[autodoc]] LlamaModel
+    - forward
+
+
+## LlamaForCausalLM [[llamaforcausallm]]
+
+[[autodoc]] LlamaForCausalLM
+    - forward
+
+## LlamaForSequenceClassification [[llamaforsequenceclassification]]
+
+[[autodoc]] LlamaForSequenceClassification
+    - forward

docs/source/ko/model_doc/llama2.md

@@ -0,0 +1,129 @@
+<!--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.
+
+⚠️ Note that this file is in Markdown but contains specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+# Llama2 [[llama2]]
+
+## 개요 [[overview]]
+
+Llama2 모델은 Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Ya1smine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, Wenyin Fu, Brian Fuller, Cynthia Gao, Vedanuj Goswami, Naman Goyal, Anthony Hartshorn, Saghar Hosseini, Rui Hou, Hakan Inan, Marcin Kardas, Viktor Kerkez Madian Khabsa, Isabel Kloumann, Artem Korenev, Punit Singh Koura, Marie-Anne Lachaux, Thibaut Lavril, Jenya Lee, Diana Liskovich, Yinghai Lu, Yuning Mao, Xavier Martinet, Todor Mihaylov, Pushkar Mishra, Igor Molybog, Yixin Nie, Andrew Poulton, Jeremy Reizenstein, Rashi Rungta, Kalyan Saladi, Alan Schelten, Ruan Silva, Eric Michael Smith, Ranjan Subramanian, Xiaoqing EllenTan, Binh Tang, Ross Taylor, Adina Williams, Jian Xiang Kuan, Puxin Xu, Zheng Yan, Iliyan Zarov, Yuchen Zhang, Angela Fan, Melanie Kambadur, Sharan Narang, Aurelien Rodriguez, Robert Stojnic, Sergey Edunov, Thomas Scialom의 논문 [LLaMA: Open Foundation and Fine-Tuned Chat Models](https://ai.meta.com/research/publications/llama-2-open-foundation-and-fine-tuned-chat-models/)에서 제안되었습니다. 채팅 어플리케이션에 맞게 미세 조정된 체크포인트를 포함된 7B에서 70B 범위의 매개변수를 가진 기초 언어 모델 모음입니다!
+
+논문의 초록은 다음과 같습니다:
+
+*이 연구에서 우리는 70억에서 700억 파라미터의 범위에서 사전 훈련 및 미세 조정된 대규모 언어 모델(LLMs)의 모음인 Llama 2를 개발 및 공개합니다. Llama 2-Chat라고 불리는 미세 조정된 LLMs은 대화 사용 사례에 최적화되었습니다. 우리의 모델은 테스트한 대부분의 벤치마크에서 오픈 소스 채팅 모델보다 성능이 뛰어나며, 유용성과 안전성에 대한 인적 평가를 바탕으로 비공개 소스 모델을 대체할 수 있는 적절한 대안이 될 수 있습니다. 우리는 Llama 2-Chat의 미세 조정 및 안전성 향상의 접근 방식에 대한 자세한 설명을 제공하여 커뮤니티가 우리의 작업을 기반으로 LLMs의 책임있는 개발에 기여할 수 있도록 합니다.*
+
+[여기](https://huggingface.co/models?search=llama2)에서 모든 Llama2 모델을 확인할 수 있습니다.
+
+<Tip warning={true}>
+
+`Llama2` 모델은 `bfloat16`을 사용하여 훈련되었지만, 원래 추론은 `float16`을 사용합니다. 허브에 업로드된 체크포인트는 `torch_dtype = 'float16'`을 사용하며, 이는 `AutoModel` API에 의해 체크포인트를 `torch.float32`에서 `torch.float16`으로 캐스팅하는 데 사용됩니다. 
+
+온라인 가중치의 `dtype`은 `model = AutoModelForCausalLM.from_pretrained("path", torch_dtype = "auto")`를 사용하여 모델을 초기화할 때 `torch_dtype="auto"`를 사용하지 않는 한 대부분 관련이 없습니다. 그 이유는 모델이 먼저 다운로드될 것이고 (온라인 체크포인트의 `dtype`을 사용하여) 그다음에 기본 `dtype`인 `torch`로 캐스팅하고(`torch.float32`가 됨), 마지막으로 구성(configuration)에서 제공된 `torch_dtype`이 있는 경우 이를 사용하기 때문입니다.
+
+모델을 `float16`에서 훈련하는 것은 권장되지 않으며 `nan`을 생성하는 것으로 알려져 있습니다. 따라서 모델은 `bfloat16`에서 훈련되어야 합니다.
+
+</Tip>
+
+🍯 팁:
+
+- Llama2 모델의 가중치는 [이 양식](https://ai.meta.com/resources/models-and-libraries/llama-downloads/)을 작성하여 얻을 수 있습니다.
+- 아키텍처는 처음 버전의 Llama와 매우 유사하며, [이 논문](https://arxiv.org/pdf/2305.13245.pdf)의 내용에 따라 Grouped Query Attention (GQA)이 추가되었습니다.
+- `config.pretraining_tp`를 1과 다른 값으로 설정하면 더 정확하지만 느린 선형 레이어 계산이 활성화되어 원본 로짓과 더 잘 일치하게 됩니다.
+- 원래 모델은 `pad_id = -1`을 사용하는데, 이는 패딩 토큰이 없음을 의미합니다. 동일한 로직을 사용할 수 없으므로 `tokenizer.add_special_tokens({"pad_token":"<pad>"})`를 사용하여 패딩 토큰을 추가하고 이에 따라 토큰 임베딩 크기를 조정해야 합니다. 또한 `model.config.pad_token_id`를 설정해야 합니다. 모델의 `embed_tokens` 레이어는 `self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.config.padding_idx)`로 초기화되어, 패딩 토큰 인코딩이 0을 출력하도록 합니다. 따라서 초기화 시에 전달하는 것을 권장합니다.
+- 양식을 작성하고 모델 체크포인트 접근 권한을 얻은 후에는 이미 변환된 체크포인트를 사용할 수 있습니다. 그렇지 않고 자신의 모델을 직접 변환하려는 경우, [변환 스크립트](https://github.com/huggingface/transformers/blob/main/src/transformers/models/llama/convert_llama_weights_to_hf.py)를 자유롭게 사용하세요. 스크립트는 다음과 같은 예시의 명령어로 호출할 수 있습니다:
+
+```bash
+python src/transformers/models/llama/convert_llama_weights_to_hf.py \
+    --input_dir /path/to/downloaded/llama/weights --model_size 7B --output_dir /output/path
+```
+
+- 변환 후 모델과 토크나이저는 다음과 같이 로드할 수 있습니다:
+
+```python
+from transformers import LlamaForCausalLM, LlamaTokenizer
+
+tokenizer = LlamaTokenizer.from_pretrained("/output/path")
+model = LlamaForCausalLM.from_pretrained("/output/path")
+```
+
+스크립트를 실행하려면 모델을 float16 정밀도로 전부 호스트할 수 있을 만큼 충분한 CPU RAM이 필요합니다 (가장 큰 버전이 여러 체크포인트로 제공되더라도 각 체크포인트는 모델 가중치의 일부만을 포함하므로 모두 RAM에 로드해야 합니다). 75B 모델의 경우, 총 145GB의 RAM이 필요합니다.
+
+- LLaMA 토크나이저는 [sentencepiece](https://github.com/google/sentencepiece)를 기반으로 한 BPE 모델입니다. sentencepiece의 특징 중 하나는 시퀀스를 디코딩할 때 첫 번째 토큰이 단어의 시작이면 (예: "Banana") 토크나이저는 문자열 앞에 접두사 공간을 추가하지 않는 것입니다.
+
+이 모델은 [Arthur Zucker](https://huggingface.co/ArthurZ)가 [Lysandre Debut](https://huggingface.co/lysandre)의 도움을 받아 제공하였습니다. Hugging Face에서의 구현 코드는 [여기](https://github.com/EleutherAI/gpt-neox)의 GPT-NeoX 를 기반으로 합니다. 저자의 원래 코드는 [여기](https://github.com/facebookresearch/llama)에서 찾을 수 있습니다.
+
+## 리소스 [[resources]]
+
+LLaMA2를 시작하는 데 도움이 될 Hugging Face의 공식 및 커뮤니티(🌎로 표시) 리소스 목록입니다. 여기에 새로운 리소스를 추가하기 위해서 Pull Request를 열어 주시면 검토하겠습니다! 리소스는 기존 리소스와 중복되지 않는 새로운 것을 보여주는 것이 이상적입니다.
+
+- [Llama 2 is here - get it on Hugging Face](https://huggingface.co/blog/llama2), Llama 2에 관한 블로그 포스트와 🤗 Transformers 및 🤗 PEFT와 함께 사용하는 방법에 대한 내용입니다.
+- [LLaMA 2 - Every Resource you need](https://www.philschmid.de/llama-2), LLaMA 2에 대해 알아보고 빠르게 시작하는 데 필요한 관련 리소스의 모음입니다.
+
+<PipelineTag pipeline="text-generation"/>
+
+- Google Colab에서 QLoRA와 4-bit 정밀도를 사용하여 Llama 2를 미세 조정하는 방법에 대한 [노트북](https://colab.research.google.com/drive/1PEQyJO1-f6j0S_XJ8DV50NkpzasXkrzd?usp=sharing)입니다. 🌎
+- "Llama-v2-7b-guanaco" 모델을 4-bit QLoRA로 미세 조정하고 PDF에서 Q&A 데이터셋을 생성하는 방법에 대한 [노트북](https://colab.research.google.com/drive/134o_cXcMe_lsvl15ZE_4Y75Kstepsntu?usp=sharing)입니다. 🌎
+
+⚗️ 최적화
+- [Llama 2를 DPO로 미세 조정하기](https://huggingface.co/blog/dpo-trl), TRL 라이브러리의 DPO 방법을 사용하여 특정 데이터셋에서 Llama 2를 미세 조정하는 방법을 안내하는 가이드입니다.
+- [확장 가이드: Llama 2 명령어 조정](https://www.philschmid.de/instruction-tune-llama-2), 입력에서 명령어를 생성하도록 Llama 2를 훈련시키는 방법을 안내하는 가이드로, 명령어를 따르는 모델에서 명령어를 주는 모델로 변환합니다.
+- 개인 컴퓨터에서 QLoRA와 TRL을 사용하여 Llama 2 모델을 미세 조정하는 방법에 대한 [노트북](https://colab.research.google.com/drive/1SYpgFpcmtIUzdE7pxqknrM4ArCASfkFQ?usp=sharing)입니다. 🌎
+
+⚡️ 추론
+- AutoGPTQ 라이브러리의 GPTQ를 사용하여 Llama 2 모델을 양자화하는 방법에 대한 [노트북](https://colab.research.google.com/drive/1TC56ArKerXUpbgRy5vM3woRsbTEVNq7h?usp=sharing)입니다. 🌎
+- 로컬 컴퓨터나 Google Colab에서 4-bit 양자화로 Llama 2 채팅 모델을 실행하는 방법에 대한 [노트북](https://colab.research.google.com/drive/1X1z9Q6domMKl2CnEM0QGHNwidLfR4dW2?usp=sharing)입니다. 🌎
+
+🚀 배포
+- [Amazon SageMaker에서 LLaMA 2 (7-70B) 미세 조정하기](https://www.philschmid.de/sagemaker-llama2-qlora), Amazon SageMaker에서 QLoRA 미세 조정 및 배포에 이르기까지의 완전한 가이드입니다.
+- [Amazon SageMaker에서 Llama 2 7B/13B/70B 배포하기](https://www.philschmid.de/sagemaker-llama-llm), 안전하고 확장 가능한 배포를 위해 Hugging Face의 LLM DLC 컨테이너를 사용하는 방법에 대한 가이드입니다.
+
+
+## LlamaConfig [[llamaconfig]]
+
+[[autodoc]] LlamaConfig
+
+
+## LlamaTokenizer [[llamatokenizer]]
+
+[[autodoc]] LlamaTokenizer
+    - build_inputs_with_special_tokens
+    - get_special_tokens_mask
+    - create_token_type_ids_from_sequences
+    - save_vocabulary
+
+## LlamaTokenizerFast [[llamatokenizerfast]]
+
+[[autodoc]] LlamaTokenizerFast
+    - build_inputs_with_special_tokens
+    - get_special_tokens_mask
+    - create_token_type_ids_from_sequences
+    - update_post_processor
+    - save_vocabulary
+
+## LlamaModel [[llamamodel]]
+
+[[autodoc]] LlamaModel
+    - forward
+
+
+## LlamaForCausalLM [[llamaforcausallm]]
+
+[[autodoc]] LlamaForCausalLM
+    - forward
+
+## LlamaForSequenceClassification [[llamaforsequenceclassification]]
+
+[[autodoc]] LlamaForSequenceClassification
+    - forward

docs/source/ms/index.md

@@ -76,6 +76,7 @@ Dokumentasi disusun kepada lima bahagian:
 1. **[BLOOM](model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](model_doc/bort)** (from Alexa) released with the paper [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) by Adrian de Wynter and Daniel J. Perry.
 1. **[BridgeTower](model_doc/bridgetower)** (from Harbin Institute of Technology/Microsoft Research Asia/Intel Labs) released with the paper [BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning](https://arxiv.org/abs/2206.08657) by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
+1. **[Bros](model_doc/bros)** (from NAVER) released with the paper [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://arxiv.org/abs/2108.04539) by Teakgyu Hong, Donghyun Kim, Mingi Ji, Wonseok Hwang, Daehyun Nam, Sungrae Park.
 1. **[ByT5](model_doc/byt5)** (from Google Research) released with the paper [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) by Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel.
 1. **[CamemBERT](model_doc/camembert)** (from Inria/Facebook/Sorbonne) released with the paper [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
 1. **[CANINE](model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
@@ -288,6 +289,7 @@ Flax), PyTorch, dan/atau TensorFlow.
 |            BLIP-2             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             BLOOM             |       ❌       |       ✅       |       ✅        |         ❌         |      ❌      |
 |          BridgeTower          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|             Bros              |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
 |           CamemBERT           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |            CANINE             |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |         Chinese-CLIP          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |

docs/source/pt/tasks/sequence_classification.md

@@ -70,7 +70,7 @@ Crie uma função de pré-processamento para tokenizar o campo `text` e truncar
 ...     return tokenizer(examples["text"], truncation=True)
 ```
 
-Use a função [`map`](https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map) do 🤗 Datasets para aplicar a função de pré-processamento em todo o conjunto de dados. Você pode acelerar a função `map` definindo `batched=True` para processar vários elementos do conjunto de dados de uma só vez:
+Use a função [`map`](https://huggingface.co/docs/datasets/process#map) do 🤗 Datasets para aplicar a função de pré-processamento em todo o conjunto de dados. Você pode acelerar a função `map` definindo `batched=True` para processar vários elementos do conjunto de dados de uma só vez:
 
 ```py
 tokenized_imdb = imdb.map(preprocess_function, batched=True)

docs/source/pt/tasks/token_classification.md

@@ -128,7 +128,7 @@ Aqui está como você pode criar uma função para realinhar os tokens e rótulo
 ...     return tokenized_inputs
 ```
 
-Use a função [`map`](https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map) do 🤗 Datasets para tokenizar e alinhar os rótulos em todo o conjunto de dados. Você pode acelerar a função `map` configurando `batched=True` para processar vários elementos do conjunto de dados de uma só vez:
+Use a função [`map`](https://huggingface.co/docs/datasets/process#map) do 🤗 Datasets para tokenizar e alinhar os rótulos em todo o conjunto de dados. Você pode acelerar a função `map` configurando `batched=True` para processar vários elementos do conjunto de dados de uma só vez:
 
 ```py
 >>> tokenized_wnut = wnut.map(tokenize_and_align_labels, batched=True)

docs/source/zh/quicktour.md

@@ -407,7 +407,7 @@ tensor([[0.0021, 0.0018, 0.0115, 0.2121, 0.7725],
 
 ## Trainer - PyTorch优化训练循环
 
-所有的模型都是标准的[`torch.nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module), 所以你可以在任何典型的训练模型中使用它们. 当你编写自己的训练循环时W, 🤗 Transformers为PyTorch提供了一个[`Trainer`]类, 它包含了基础的训练循环并且为诸如分布式训练, 混合精度等特性增加了额外的功能.
+所有的模型都是标准的[`torch.nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module), 所以你可以在任何典型的训练模型中使用它们. 当你编写自己的训练循环时, 🤗 Transformers为PyTorch提供了一个[`Trainer`]类, 它包含了基础的训练循环并且为诸如分布式训练, 混合精度等特性增加了额外的功能.
 
 取决于你的任务, 你通常可以传递以下的参数给[`Trainer`]:
 

examples/README.md

@@ -19,7 +19,7 @@ We host a wide range of example scripts for multiple learning frameworks. Simply
 
 We also have some [research projects](https://github.com/huggingface/transformers/tree/main/examples/research_projects), as well as some [legacy examples](https://github.com/huggingface/transformers/tree/main/examples/legacy). Note that unlike the main examples these are not actively maintained, and may require specific older versions of dependencies in order to run. 
 
-While we strive to present as many use cases as possible, the example scripts are just that - examples. It is expected that they won't work out-of-the box on your specific problem and that you will be required to change a few lines of code to adapt them to your needs. To help you with that, most of the examples fully expose the preprocessing of the data, allowing you to tweak and edit them as required.
+While we strive to present as many use cases as possible, the example scripts are just that - examples. It is expected that they won't work out-of-the-box on your specific problem and that you will be required to change a few lines of code to adapt them to your needs. To help you with that, most of the examples fully expose the preprocessing of the data, allowing you to tweak and edit them as required.
 
 Please discuss on the [forum](https://discuss.huggingface.co/) or in an [issue](https://github.com/huggingface/transformers/issues) a feature you would like to implement in an example before submitting a PR; we welcome bug fixes, but since we want to keep the examples as simple as possible it's unlikely that we will merge a pull request adding more functionality at the cost of readability.
 

examples/flax/language-modeling/run_bart_dlm_flax.py

@@ -139,7 +139,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )
@@ -684,7 +684,7 @@ def main():
     # might be slower to preprocess.
     #
     # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-    # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+    # https://huggingface.co/docs/datasets/process#map
     tokenized_datasets = tokenized_datasets.map(
         group_texts,
         batched=True,

examples/flax/language-modeling/run_clm_flax.py

@@ -140,7 +140,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )
@@ -607,7 +607,7 @@ def main():
     # to preprocess.
     #
     # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-    # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+    # https://huggingface.co/docs/datasets/process#map
 
     lm_datasets = tokenized_datasets.map(
         group_texts,

examples/flax/language-modeling/run_mlm_flax.py

@@ -145,7 +145,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )
@@ -625,7 +625,7 @@ def main():
         # might be slower to preprocess.
         #
         # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-        # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+        # https://huggingface.co/docs/datasets/process#map
         tokenized_datasets = tokenized_datasets.map(
             group_texts,
             batched=True,

examples/flax/language-modeling/run_t5_mlm_flax.py

@@ -139,7 +139,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )
@@ -715,7 +715,7 @@ def main():
     # might be slower to preprocess.
     #
     # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-    # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+    # https://huggingface.co/docs/datasets/process#map
     tokenized_datasets = tokenized_datasets.map(
         group_texts,
         batched=True,

examples/flax/question-answering/run_qa.py

@@ -62,7 +62,7 @@ from transformers.utils import check_min_version, send_example_telemetry
 logger = logging.getLogger(__name__)
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 Array = Any
 Dataset = datasets.arrow_dataset.Dataset

examples/flax/summarization/run_summarization_flax.py

@@ -159,7 +159,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )

examples/flax/text-classification/run_flax_glue.py

@@ -55,7 +55,7 @@ from transformers.utils import check_min_version, send_example_telemetry
 
 logger = logging.getLogger(__name__)
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 Array = Any
 Dataset = datasets.arrow_dataset.Dataset

examples/flax/token-classification/run_flax_ner.py

@@ -56,7 +56,7 @@ from transformers.utils.versions import require_version
 
 logger = logging.getLogger(__name__)
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/token-classification/requirements.txt")
 

examples/flax/vision/run_image_classification.py

@@ -137,7 +137,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )

examples/legacy/seq2seq/seq2seq_training_args.py

@@ -31,7 +31,7 @@ class Seq2SeqTrainingArguments(TrainingArguments):
         label_smoothing (:obj:`float`, `optional`, defaults to 0):
             The label smoothing epsilon to apply (if not zero).
         sortish_sampler (:obj:`bool`, `optional`, defaults to :obj:`False`):
-            Whether to SortishSamler or not. It sorts the inputs according to lenghts in-order to minimizing the padding size.
+            Whether to SortishSamler or not. It sorts the inputs according to lengths in-order to minimizing the padding size.
         predict_with_generate (:obj:`bool`, `optional`, defaults to :obj:`False`):
             Whether to use generate to calculate generative metrics (ROUGE, BLEU).
     """

examples/pytorch/audio-classification/run_audio_classification.py

@@ -45,7 +45,7 @@ from transformers.utils.versions import require_version
 logger = logging.getLogger(__name__)
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.14.0", "To fix: pip install -r examples/pytorch/audio-classification/requirements.txt")
 

examples/pytorch/contrastive-image-text/run_clip.py

@@ -55,7 +55,7 @@ from transformers.utils.versions import require_version
 logger = logging.getLogger(__name__)
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/contrastive-image-text/requirements.txt")
 

examples/pytorch/image-classification/run_image_classification.py

@@ -56,7 +56,7 @@ from transformers.utils.versions import require_version
 logger = logging.getLogger(__name__)
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/image-classification/requirements.txt")
 

examples/pytorch/image-classification/run_image_classification_no_trainer.py

@@ -47,7 +47,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -477,8 +477,8 @@ def main():
             # need to multiply `gradient_accumulation_steps` to reflect real steps
             resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
             starting_epoch = resume_step // len(train_dataloader)
+            completed_steps = resume_step // args.gradient_accumulation_steps
             resume_step -= starting_epoch * len(train_dataloader)
-            completed_steps = resume_step // args.gradient_accumulation_step
 
     # update the progress_bar if load from checkpoint
     progress_bar.update(completed_steps)
@@ -511,7 +511,7 @@ def main():
 
             if isinstance(checkpointing_steps, int):
                 if completed_steps % checkpointing_steps == 0:
-                    output_dir = f"step_{completed_steps }"
+                    output_dir = f"step_{completed_steps}"
                     if args.output_dir is not None:
                         output_dir = os.path.join(args.output_dir, output_dir)
                     accelerator.save_state(output_dir)

examples/pytorch/image-pretraining/run_mae.py

@@ -44,7 +44,7 @@ from transformers.utils.versions import require_version
 logger = logging.getLogger(__name__)
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/image-pretraining/requirements.txt")
 
@@ -110,7 +110,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )
@@ -163,15 +163,6 @@ class CustomTrainingArguments(TrainingArguments):
         default=1e-3, metadata={"help": "Base learning rate: absolute_lr = base_lr * total_batch_size / 256."}
     )
 
-    def __post_init__(self):
-        # Compute absolute learning rate while args are mutable
-        super().__post_init__()
-        if self.base_learning_rate is not None:
-            total_train_batch_size = self.train_batch_size * self.gradient_accumulation_steps * self.world_size
-            delattr(self, "_frozen")
-            self.learning_rate = self.base_learning_rate * total_train_batch_size / 256
-            setattr(self, "_frozen", True)
-
 
 def collate_fn(examples):
     pixel_values = torch.stack([example["pixel_values"] for example in examples])
@@ -362,6 +353,13 @@ def main():
         # Set the validation transforms
         ds["validation"].set_transform(preprocess_images)
 
+    # Compute absolute learning rate
+    total_train_batch_size = (
+        training_args.train_batch_size * training_args.gradient_accumulation_steps * training_args.world_size
+    )
+    if training_args.base_learning_rate is not None:
+        training_args.learning_rate = training_args.base_learning_rate * total_train_batch_size / 256
+
     # Initialize our trainer
     trainer = Trainer(
         model=model,

examples/pytorch/image-pretraining/run_mim.py

@@ -49,7 +49,7 @@ Any model supported by the AutoModelForMaskedImageModeling API can be used.
 logger = logging.getLogger(__name__)
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/image-pretraining/requirements.txt")
 

examples/pytorch/image-pretraining/run_mim_no_trainer.py

@@ -54,7 +54,7 @@ Any model supported by the AutoModelForMaskedImageModeling API can be used.
 logger = logging.getLogger(__name__)
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/image-pretraining/requirements.txt")
 
@@ -701,8 +701,8 @@ def main():
             # need to multiply `gradient_accumulation_steps` to reflect real steps
             resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
             starting_epoch = resume_step // len(train_dataloader)
-            resume_step -= starting_epoch * len(train_dataloader)
             completed_steps = resume_step // args.gradient_accumulation_steps
+            resume_step -= starting_epoch * len(train_dataloader)
 
     # update the progress_bar if load from checkpoint
     progress_bar.update(completed_steps)
@@ -735,7 +735,7 @@ def main():
 
             if isinstance(checkpointing_steps, int):
                 if completed_steps % checkpointing_steps == 0:
-                    output_dir = f"step_{completed_steps }"
+                    output_dir = f"step_{completed_steps}"
                     if args.output_dir is not None:
                         output_dir = os.path.join(args.output_dir, output_dir)
                     accelerator.save_state(output_dir)

examples/pytorch/language-modeling/run_clm.py

@@ -56,7 +56,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/language-modeling/requirements.txt")
 
@@ -77,7 +77,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )
@@ -533,7 +533,7 @@ def main():
     # to preprocess.
     #
     # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-    # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+    # https://huggingface.co/docs/datasets/process#map
 
     with training_args.main_process_first(desc="grouping texts together"):
         if not data_args.streaming:

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -57,7 +57,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -473,7 +473,7 @@ def main():
     # to preprocess.
     #
     # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-    # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+    # https://huggingface.co/docs/datasets/process#map
 
     with accelerator.main_process_first():
         lm_datasets = tokenized_datasets.map(
@@ -632,7 +632,7 @@ def main():
 
             if isinstance(checkpointing_steps, int):
                 if completed_steps % checkpointing_steps == 0:
-                    output_dir = f"step_{completed_steps }"
+                    output_dir = f"step_{completed_steps}"
                     if args.output_dir is not None:
                         output_dir = os.path.join(args.output_dir, output_dir)
                     accelerator.save_state(output_dir)

examples/pytorch/language-modeling/run_mlm.py

@@ -54,7 +54,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/language-modeling/requirements.txt")
 
@@ -547,7 +547,7 @@ def main():
         # might be slower to preprocess.
         #
         # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-        # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+        # https://huggingface.co/docs/datasets/process#map
 
         with training_args.main_process_first(desc="grouping texts together"):
             if not data_args.streaming:

examples/pytorch/language-modeling/run_mlm_no_trainer.py

@@ -57,7 +57,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 logger = get_logger(__name__)
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/language-modeling/requirements.txt")
@@ -504,7 +504,7 @@ def main():
         # might be slower to preprocess.
         #
         # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-        # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+        # https://huggingface.co/docs/datasets/process#map
 
         with accelerator.main_process_first():
             tokenized_datasets = tokenized_datasets.map(
@@ -636,8 +636,8 @@ def main():
             # need to multiply `gradient_accumulation_steps` to reflect real steps
             resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
             starting_epoch = resume_step // len(train_dataloader)
-            resume_step -= starting_epoch * len(train_dataloader)
             completed_steps = resume_step // args.gradient_accumulation_steps
+            resume_step -= starting_epoch * len(train_dataloader)
 
     # update the progress_bar if load from checkpoint
     progress_bar.update(completed_steps)
@@ -670,7 +670,7 @@ def main():
 
             if isinstance(checkpointing_steps, int):
                 if completed_steps % checkpointing_steps == 0:
-                    output_dir = f"step_{completed_steps }"
+                    output_dir = f"step_{completed_steps}"
                     if args.output_dir is not None:
                         output_dir = os.path.join(args.output_dir, output_dir)
                     accelerator.save_state(output_dir)

examples/pytorch/language-modeling/run_plm.py

@@ -48,7 +48,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/language-modeling/requirements.txt")
 
@@ -65,7 +65,7 @@ class ModelArguments:
         default=None,
         metadata={
             "help": (
-                "The model checkpoint for weights initialization.Don't set if you want to train a model from scratch."
+                "The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
             )
         },
     )
@@ -478,7 +478,7 @@ def main():
         # might be slower to preprocess.
         #
         # To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
-        # https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
+        # https://huggingface.co/docs/datasets/process#map
 
         with training_args.main_process_first(desc="grouping texts together"):
             tokenized_datasets = tokenized_datasets.map(

examples/pytorch/multiple-choice/run_swag.py

@@ -48,7 +48,7 @@ from transformers.utils import PaddingStrategy, check_min_version, send_example_
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 logger = logging.getLogger(__name__)
 

examples/pytorch/multiple-choice/run_swag_no_trainer.py

@@ -56,7 +56,7 @@ from transformers.utils import PaddingStrategy, check_min_version, send_example_
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 logger = get_logger(__name__)
 # You should update this to your particular problem to have better documentation of `model_type`
@@ -583,8 +583,8 @@ def main():
             # need to multiply `gradient_accumulation_steps` to reflect real steps
             resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
             starting_epoch = resume_step // len(train_dataloader)
+            completed_steps = resume_step // args.gradient_accumulation_steps
             resume_step -= starting_epoch * len(train_dataloader)
-            completed_steps = resume_step // args.gradient_accumulation_stepp
 
     # update the progress_bar if load from checkpoint
     progress_bar.update(completed_steps)
@@ -617,7 +617,7 @@ def main():
 
             if isinstance(checkpointing_steps, int):
                 if completed_steps % checkpointing_steps == 0:
-                    output_dir = f"step_{completed_steps }"
+                    output_dir = f"step_{completed_steps}"
                     if args.output_dir is not None:
                         output_dir = os.path.join(args.output_dir, output_dir)
                     accelerator.save_state(output_dir)

examples/pytorch/question-answering/run_qa.py

@@ -50,7 +50,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/question-answering/requirements.txt")
 

examples/pytorch/question-answering/run_qa_beam_search.py

@@ -49,7 +49,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/question-answering/requirements.txt")
 

examples/pytorch/question-answering/run_qa_beam_search_no_trainer.py

@@ -56,7 +56,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/question-answering/requirements.txt")
 
@@ -820,8 +820,8 @@ def main():
             # need to multiply `gradient_accumulation_steps` to reflect real steps
             resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
             starting_epoch = resume_step // len(train_dataloader)
+            completed_steps = resume_step // args.gradient_accumulation_steps
             resume_step -= starting_epoch * len(train_dataloader)
-            completed_steps = resume_step // args.gradient_accumulation_stepp
 
     # update the progress_bar if load from checkpoint
     progress_bar.update(completed_steps)

examples/pytorch/question-answering/run_qa_no_trainer.py

@@ -57,7 +57,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/question-answering/requirements.txt")
 
@@ -848,10 +848,11 @@ def main():
             resume_step = None
             completed_steps = starting_epoch * num_update_steps_per_epoch
         else:
-            resume_step = int(training_difference.replace("step_", ""))
+            # need to multiply `gradient_accumulation_steps` to reflect real steps
+            resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
             starting_epoch = resume_step // len(train_dataloader)
+            completed_steps = resume_step // args.gradient_accumulation_steps
             resume_step -= starting_epoch * len(train_dataloader)
-            completed_steps = resume_step // args.gradient_accumulation_stepp
 
     # update the progress_bar if load from checkpoint
     progress_bar.update(completed_steps)
@@ -885,7 +886,7 @@ def main():
 
             if isinstance(checkpointing_steps, int):
                 if completed_steps % checkpointing_steps == 0:
-                    output_dir = f"step_{completed_steps }"
+                    output_dir = f"step_{completed_steps}"
                     if args.output_dir is not None:
                         output_dir = os.path.join(args.output_dir, output_dir)
                     accelerator.save_state(output_dir)

examples/pytorch/question-answering/run_seq2seq_qa.py

@@ -47,7 +47,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/question-answering/requirements.txt")
 

examples/pytorch/question-answering/trainer_seq2seq_qa.py

@@ -46,12 +46,13 @@ class QuestionAnsweringSeq2SeqTrainer(Seq2SeqTrainer):
         **gen_kwargs,
     ) -> Dict[str, float]:
         gen_kwargs = gen_kwargs.copy()
-        gen_kwargs["max_length"] = (
-            gen_kwargs["max_length"] if gen_kwargs.get("max_length") is not None else self.args.generation_max_length
-        )
-        gen_kwargs["num_beams"] = (
-            gen_kwargs["num_beams"] if gen_kwargs.get("num_beams") is not None else self.args.generation_num_beams
-        )
+
+        # Use legacy argument setting if a) the option is not explicitly passed; and b) the argument is set in the
+        # training args
+        if gen_kwargs.get("max_length") is None and self.args.generation_max_length is not None:
+            gen_kwargs["max_length"] = self.args.generation_max_length
+        if gen_kwargs.get("num_beams") is None and self.args.generation_num_beams is not None:
+            gen_kwargs["num_beams"] = self.args.generation_num_beams
         self._gen_kwargs = gen_kwargs
 
         eval_dataset = self.eval_dataset if eval_dataset is None else eval_dataset

examples/pytorch/semantic-segmentation/run_semantic_segmentation.py

@@ -52,7 +52,7 @@ from transformers.utils.versions import require_version
 logger = logging.getLogger(__name__)
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=2.0.0", "To fix: pip install -r examples/pytorch/semantic-segmentation/requirements.txt")
 

examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py

@@ -50,7 +50,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 logger = get_logger(__name__)
 
@@ -581,8 +581,8 @@ def main():
             # need to multiply `gradient_accumulation_steps` to reflect real steps
             resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
             starting_epoch = resume_step // len(train_dataloader)
+            completed_steps = resume_step // args.gradient_accumulation_steps
             resume_step -= starting_epoch * len(train_dataloader)
-            completed_steps = resume_step // args.gradient_accumulation_stepp
 
     # update the progress_bar if load from checkpoint
     progress_bar.update(completed_steps)
@@ -615,7 +615,7 @@ def main():
 
             if isinstance(checkpointing_steps, int):
                 if completed_steps % checkpointing_steps == 0:
-                    output_dir = f"step_{completed_steps }"
+                    output_dir = f"step_{completed_steps}"
                     if args.output_dir is not None:
                         output_dir = os.path.join(args.output_dir, output_dir)
                     accelerator.save_state(output_dir)

examples/pytorch/speech-recognition/run_speech_recognition_ctc.py

@@ -51,7 +51,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.18.0", "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt")
 
@@ -311,7 +311,7 @@ class DataCollatorCTCWithPadding:
     pad_to_multiple_of_labels: Optional[int] = None
 
     def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
-        # split inputs and labels since they have to be of different lenghts and need
+        # split inputs and labels since they have to be of different lengths and need
         # different padding methods
         input_features = [{"input_values": feature["input_values"]} for feature in features]
         label_features = [{"input_ids": feature["labels"]} for feature in features]

examples/pytorch/speech-recognition/run_speech_recognition_ctc_adapter.py

@@ -53,7 +53,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.18.0", "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt")
 
@@ -307,7 +307,7 @@ class DataCollatorCTCWithPadding:
     pad_to_multiple_of_labels: Optional[int] = None
 
     def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
-        # split inputs and labels since they have to be of different lenghts and need
+        # split inputs and labels since they have to be of different lengths and need
         # different padding methods
         input_features = [{"input_values": feature["input_values"]} for feature in features]
         label_features = [{"input_ids": feature["labels"]} for feature in features]

examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py

@@ -49,7 +49,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.18.0", "To fix: pip install -r examples/pytorch/speech-recognition/requirements.txt")
 

examples/pytorch/summarization/run_summarization.py

@@ -18,7 +18,6 @@ Fine-tuning the library models for sequence to sequence.
 """
 # You can also adapt this script on your own sequence to sequence task. Pointers for this are left as comments.
 
-import dataclasses
 import logging
 import os
 import sys
@@ -54,7 +53,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/summarization/requirements.txt")
 
@@ -675,10 +674,14 @@ def main():
         return result
 
     # Override the decoding parameters of Seq2SeqTrainer
-    if training_args.generation_max_length is None:
-        training_args = dataclasses.replace(training_args, generation_max_length=data_args.val_max_target_length)
-    if training_args.generation_num_beams is None:
-        training_args = dataclasses.replace(training_args, generation_num_beams=data_args.num_beams)
+    training_args.generation_max_length = (
+        training_args.generation_max_length
+        if training_args.generation_max_length is not None
+        else data_args.val_max_target_length
+    )
+    training_args.generation_num_beams = (
+        data_args.num_beams if data_args.num_beams is not None else training_args.generation_num_beams
+    )
 
     # Initialize our Trainer
     trainer = Seq2SeqTrainer(

examples/pytorch/summarization/run_summarization_no_trainer.py

@@ -56,7 +56,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 logger = get_logger(__name__)
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/summarization/requirements.txt")
@@ -652,8 +652,8 @@ def main():
             # need to multiply `gradient_accumulation_steps` to reflect real steps
             resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
             starting_epoch = resume_step // len(train_dataloader)
+            completed_steps = resume_step // args.gradient_accumulation_steps
             resume_step -= starting_epoch * len(train_dataloader)
-            completed_steps = resume_step // args.gradient_accumulation_stepp
 
     # update the progress_bar if load from checkpoint
     progress_bar.update(completed_steps)
@@ -686,7 +686,7 @@ def main():
 
             if isinstance(checkpointing_steps, int):
                 if completed_steps % checkpointing_steps == 0:
-                    output_dir = f"step_{completed_steps }"
+                    output_dir = f"step_{completed_steps}"
                     if args.output_dir is not None:
                         output_dir = os.path.join(args.output_dir, output_dir)
                     accelerator.save_state(output_dir)

examples/pytorch/text-classification/run_classification.py

@@ -48,7 +48,7 @@ from transformers.utils.versions import require_version
 
 
 # Will error if the minimal version of Transformers is not installed. Remove at your own risks.
-check_min_version("4.33.0.dev0")
+check_min_version("4.34.0.dev0")
 
 require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")