diff --git a/README.md b/README.md index fb74507e7f..4416cfee8b 100644 --- a/README.md +++ b/README.md @@ -409,6 +409,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro. 1. **[Megatron-GPT2](https://huggingface.co/docs/transformers/model_doc/megatron_gpt2)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro. 1. **[MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str)** (from Alibaba Research) released with the paper [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592) by Peng Wang, Cheng Da, and Cong Yao. +1. **[Mistral](https://huggingface.co/docs/transformers/model_doc/mistral)** (from Mistral AI) by The [Mistral AI](https://mistral.ai) team: Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed. 1. **[mLUKE](https://huggingface.co/docs/transformers/model_doc/mluke)** (from Studio Ousia) released with the paper [mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models](https://arxiv.org/abs/2110.08151) by Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka. 1. **[MMS](https://huggingface.co/docs/transformers/model_doc/mms)** (from Facebook) released with the paper [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516) by Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, Michael Auli. 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. diff --git a/README_es.md b/README_es.md index 5af1bba38a..77054f6e98 100644 --- a/README_es.md +++ b/README_es.md @@ -385,6 +385,7 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro. 1. **[Megatron-GPT2](https://huggingface.co/docs/transformers/model_doc/megatron_gpt2)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro. 1. **[MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str)** (from Alibaba Research) released with the paper [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592) by Peng Wang, Cheng Da, and Cong Yao. +1. **[Mistral](https://huggingface.co/docs/transformers/model_doc/mistral)** (from Mistral AI) by The Mistral AI team: Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed.. 1. **[mLUKE](https://huggingface.co/docs/transformers/model_doc/mluke)** (from Studio Ousia) released with the paper [mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models](https://arxiv.org/abs/2110.08151) by Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka. 1. **[MMS](https://huggingface.co/docs/transformers/model_doc/mms)** (from Facebook) released with the paper [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516) by Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, Michael Auli. 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. diff --git a/README_hd.md b/README_hd.md index c907eaa6b1..8c279db66a 100644 --- a/README_hd.md +++ b/README_hd.md @@ -357,6 +357,7 @@ conda install -c huggingface transformers 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (NVIDIA से) कागज के साथ [Megatron-LM: मॉडल का उपयोग करके बहु-अरब पैरामीटर भाषा मॉडल का प्रशिक्षण Parallelism](https://arxiv.org/abs/1909.08053) मोहम्मद शोएबी, मोस्टोफा पटवारी, राउल पुरी, पैट्रिक लेग्रेस्ले, जेरेड कैस्पर और ब्रायन कैटानज़ारो द्वारा। 1. **[Megatron-GPT2](https://huggingface.co/docs/transformers/model_doc/megatron_gpt2)** (NVIDIA से) साथ वाला पेपर [Megatron-LM: ट्रेनिंग मल्टी-बिलियन पैरामीटर लैंग्वेज मॉडल्स यूजिंग मॉडल पैरेललिज़्म] (https://arxiv.org/abs/1909.08053) मोहम्मद शोएबी, मोस्टोफा पटवारी, राउल पुरी, पैट्रिक लेग्रेस्ले, जेरेड कैस्पर और ब्रायन कैटानज़ारो द्वारा पोस्ट किया गया। 1. **[MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str)** (Alibaba Research से) Peng Wang, Cheng Da, and Cong Yao. द्वाराअनुसंधान पत्र [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592) के साथ जारी किया गया +1. **[Mistral](https://huggingface.co/docs/transformers/model_doc/mistral)** (from Mistral AI) by The Mistral AI team: Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed.. 1. **[mLUKE](https://huggingface.co/docs/transformers/model_doc/mluke)** (फ्रॉम Studio Ousia) साथ में पेपर [mLUKE: द पावर ऑफ एंटिटी रिप्रेजेंटेशन इन मल्टीलिंगुअल प्रीट्रेन्ड लैंग्वेज मॉडल्स](https://arxiv.org/abs/2110.08151) रयोकन री, इकुया यामाडा, और योशिमासा त्सुरोका द्वारा। 1. **[MMS](https://huggingface.co/docs/transformers/model_doc/mms)** (Facebook से) Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, Michael Auli. द्वाराअनुसंधान पत्र [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516) के साथ जारी किया गया 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (सीएमयू/गूगल ब्रेन से) साथ में कागज [मोबाइलबर्ट: संसाधन-सीमित उपकरणों के लिए एक कॉम्पैक्ट टास्क-अज्ञेय बीईआरटी] (https://arxiv.org/abs/2004.02984) Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, और Denny Zhou द्वारा पोस्ट किया गया। diff --git a/README_ja.md b/README_ja.md index af8bb97702..97bb6a648f 100644 --- a/README_ja.md +++ b/README_ja.md @@ -419,6 +419,7 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (NVIDIA から) Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro から公開された研究論文: [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 1. **[Megatron-GPT2](https://huggingface.co/docs/transformers/model_doc/megatron_gpt2)** (NVIDIA から) Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro から公開された研究論文: [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 1. **[MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str)** (Alibaba Research から) Peng Wang, Cheng Da, and Cong Yao. から公開された研究論文 [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592) +1. **[Mistral](https://huggingface.co/docs/transformers/model_doc/mistral)** (from Mistral AI) by The Mistral AI team: Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed.. 1. **[mLUKE](https://huggingface.co/docs/transformers/model_doc/mluke)** (Studio Ousia から) Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka から公開された研究論文: [mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models](https://arxiv.org/abs/2110.08151) 1. **[MMS](https://huggingface.co/docs/transformers/model_doc/mms)** (Facebook から) Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, Michael Auli. から公開された研究論文 [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516) 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (CMU/Google Brain から) Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou から公開された研究論文: [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) diff --git a/README_ko.md b/README_ko.md index 43dc072827..e854439246 100644 --- a/README_ko.md +++ b/README_ko.md @@ -334,6 +334,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (NVIDIA 에서) Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro 의 [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 논문과 함께 발표했습니다. 1. **[Megatron-GPT2](https://huggingface.co/docs/transformers/model_doc/megatron_gpt2)** (NVIDIA 에서) Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro 의 [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 논문과 함께 발표했습니다. 1. **[MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str)** (Alibaba Research 에서 제공)은 Peng Wang, Cheng Da, and Cong Yao.의 [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592)논문과 함께 발표했습니다. +1. **[Mistral](https://huggingface.co/docs/transformers/model_doc/mistral)** (from Mistral AI) by The Mistral AI team: Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed.. 1. **[mLUKE](https://huggingface.co/docs/transformers/model_doc/mluke)** (Studio Ousia 에서) Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka 의 [mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models](https://arxiv.org/abs/2110.08151) 논문과 함께 발표했습니다. 1. **[MMS](https://huggingface.co/docs/transformers/model_doc/mms)** (Facebook 에서 제공)은 Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, Michael Auli.의 [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516)논문과 함께 발표했습니다. 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (CMU/Google Brain 에서) Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou 의 [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) 논문과 함께 발표했습니다. diff --git a/README_zh-hans.md b/README_zh-hans.md index eee688a258..8fa9212c1f 100644 --- a/README_zh-hans.md +++ b/README_zh-hans.md @@ -358,6 +358,7 @@ conda install -c huggingface transformers 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (来自 NVIDIA) 伴随论文 [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 由 Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro 发布。 1. **[Megatron-GPT2](https://huggingface.co/docs/transformers/model_doc/megatron_gpt2)** (来自 NVIDIA) 伴随论文 [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 由 Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro 发布。 1. **[MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str)** (来自 Alibaba Research) 伴随论文 [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592) 由 Peng Wang, Cheng Da, and Cong Yao 发布。 +1. **[Mistral](https://huggingface.co/docs/transformers/model_doc/mistral)** (from Mistral AI) by The Mistral AI team: Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed.. 1. **[mLUKE](https://huggingface.co/docs/transformers/model_doc/mluke)** (来自 Studio Ousia) 伴随论文 [mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models](https://arxiv.org/abs/2110.08151) 由 Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka 发布。 1. **[MMS](https://huggingface.co/docs/transformers/model_doc/mms)** (来自 Facebook) 伴随论文 [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516) 由 Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, Michael Auli 发布。 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (来自 CMU/Google Brain) 伴随论文 [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) 由 Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou 发布。 diff --git a/README_zh-hant.md b/README_zh-hant.md index f939bb8e37..83254ab0c0 100644 --- a/README_zh-hant.md +++ b/README_zh-hant.md @@ -370,6 +370,7 @@ conda install -c huggingface transformers 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro. 1. **[Megatron-GPT2](https://huggingface.co/docs/transformers/model_doc/megatron_gpt2)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro. 1. **[MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str)** (from Alibaba Research) released with the paper [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592) by Peng Wang, Cheng Da, and Cong Yao. +1. **[Mistral](https://huggingface.co/docs/transformers/model_doc/mistral)** (from Mistral AI) by The Mistral AI team: Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed.. 1. **[mLUKE](https://huggingface.co/docs/transformers/model_doc/mluke)** (from Studio Ousia) released with the paper [mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models](https://arxiv.org/abs/2110.08151) by Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka. 1. **[MMS](https://huggingface.co/docs/transformers/model_doc/mms)** (from Facebook) released with the paper [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516) by Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, Michael Auli. 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. diff --git a/docs/source/en/_toctree.yml b/docs/source/en/_toctree.yml index de9fe55d95..b8172071b8 100644 --- a/docs/source/en/_toctree.yml +++ b/docs/source/en/_toctree.yml @@ -384,6 +384,8 @@ title: MegatronBERT - local: model_doc/megatron_gpt2 title: MegatronGPT2 + - local: model_doc/mistral + title: Mistral - local: model_doc/mluke title: mLUKE - local: model_doc/mobilebert diff --git a/docs/source/en/index.md b/docs/source/en/index.md index 49370b12fa..217ae800da 100644 --- a/docs/source/en/index.md +++ b/docs/source/en/index.md @@ -174,6 +174,7 @@ The documentation is organized into five sections: 1. **[Megatron-BERT](model_doc/megatron-bert)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro. 1. **[Megatron-GPT2](model_doc/megatron_gpt2)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro. 1. **[MGP-STR](model_doc/mgp-str)** (from Alibaba Research) released with the paper [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592) by Peng Wang, Cheng Da, and Cong Yao. +1. **[Mistral](model_doc/mistral)** (from Mistral AI) by The [Mistral AI](https://mistral.ai) team: Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed. 1. **[mLUKE](model_doc/mluke)** (from Studio Ousia) released with the paper [mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models](https://arxiv.org/abs/2110.08151) by Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka. 1. **[MMS](model_doc/mms)** (from Facebook) released with the paper [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516) by Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, Michael Auli. 1. **[MobileBERT](model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. @@ -398,6 +399,7 @@ Flax), PyTorch, and/or TensorFlow. | MEGA | ✅ | ❌ | ❌ | | Megatron-BERT | ✅ | ❌ | ❌ | | MGP-STR | ✅ | ❌ | ❌ | +| Mistral | ✅ | ❌ | ❌ | | MobileBERT | ✅ | ✅ | ❌ | | MobileNetV1 | ✅ | ❌ | ❌ | | MobileNetV2 | ✅ | ❌ | ❌ | diff --git a/docs/source/en/model_doc/mistral.md b/docs/source/en/model_doc/mistral.md new file mode 100644 index 0000000000..578906abe6 --- /dev/null +++ b/docs/source/en/model_doc/mistral.md @@ -0,0 +1,106 @@ + + +# Mistral + +## Overview + +Mistral-7B-v0.1 is Mistral AI’s first Large Language Model (LLM). + +## Model Details + +Mistral-7B-v0.1 is a decoder-based LM with the following architectural choices: +* Sliding Window Attention - Trained with 8k context length and fixed cache size, with a theoretical attention span of 128K tokens +* GQA (Grouped Query Attention) - allowing faster inference and lower cache size. +* Byte-fallback BPE tokenizer - ensures that characters are never mapped to out of vocabulary tokens. + +We also provide an instruction fine-tuned model: `Mistral-7B-Instruct-v0.1` which can be used for chat-based inference. + +For more details please read our [release blog post](https://mistral.ai/news/announcing-mistral-7b-v0.1/) + +## License + +Both `Mistral-7B-v0.1` and `Mistral-7B-Instruct-v0.1` are released under the Apache 2.0 license. + +## Usage + +`Mistral-7B-v0.1` and `Mistral-7B-Instruct-v0.1` can be found on the [Huggingface Hub](https://huggingface.co/mistralai) + +These ready-to-use checkpoints can be downloaded and used via the HuggingFace Hub: + +```python +>>> from transformers import AutoModelForCausalLM, AutoTokenizer +>>> device = "cuda" # the device to load the model onto + +>>> model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1") +>>> tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-v0.1") + +>>> prompt = "My favourite condiment is" + +>>> model_inputs = tokenizer([prompt], return_tensors="pt").to(device) +>>> model.to(device) + +>>> generated_ids = model.generate(**model_inputs, max_new_tokens=100, do_sample=True) +>>> tokenizer.batch_decode(generated_ids)[0] +"The expected outupt" +``` + +Raw weights for `Mistral-7B-v0.1` and `Mistral-7B-Instruct-v0.1` can be downloaded from: + +| Model Name | Checkpoint | +|----------------------------|-----------------------------------------------------------------------------------------| +| `Mistral-7B-v0.1` | [Raw Checkpoint](https://files.mistral-7b-v0-1.mistral.ai/mistral-7B-v0.1.tar) | +| `Mistral-7B-Instruct-v0.1` | [Raw Checkpoint](https://files.mistral-7b-v0-1.mistral.ai/mistral-7B-instruct-v0.1.tar) | + + +To use these raw checkpoints with HuggingFace you can use the `convert_mistral_weights_to_hf.py` script to convert them to the HuggingFace format: + +```bash +python src/transformers/models/mistral/convert_mistral_weights_to_hf.py \ + --input_dir /path/to/downloaded/mistral/weights --model_size 7B --output_dir /output/path +``` + +You can then load the converted model from the `output/path`: + +```python +from transformers import MistralForCausalLM, LlamaTokenzier + +tokenizer = LlamaTokenizer.from_pretrained("/output/path") +model = MistralForCausalLM.from_pretrained("/output/path") +``` + +## The Mistral Team + +Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed. + +## MistralConfig + +[[autodoc]] MistralConfig + +## MistralModel + +[[autodoc]] MistralModel + - forward + +## MistralForCausalLM + +[[autodoc]] MistralForCausalLM + - forward + +## MistralForSequenceClassification + +[[autodoc]] MistralForSequenceClassification + - forward \ No newline at end of file diff --git a/docs/source/en/tasks/language_modeling.md b/docs/source/en/tasks/language_modeling.md index 71de42151e..4240678821 100644 --- a/docs/source/en/tasks/language_modeling.md +++ b/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: -[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) +[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), [Mistral](../model_doc/mistral), [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) diff --git a/docs/source/en/tasks/sequence_classification.md b/docs/source/en/tasks/sequence_classification.md index ec0509bd60..b67d43453d 100644 --- a/docs/source/en/tasks/sequence_classification.md +++ b/docs/source/en/tasks/sequence_classification.md @@ -33,7 +33,7 @@ The task illustrated in this tutorial is supported by the following model archit -[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) +[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), [Mistral](../model_doc/mistral), [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) diff --git a/src/transformers/__init__.py b/src/transformers/__init__.py index 84aca87ded..c093762b3c 100644 --- a/src/transformers/__init__.py +++ b/src/transformers/__init__.py @@ -433,6 +433,7 @@ _import_structure = { "models.megatron_bert": ["MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MegatronBertConfig"], "models.megatron_gpt2": [], "models.mgp_str": ["MGP_STR_PRETRAINED_CONFIG_ARCHIVE_MAP", "MgpstrConfig", "MgpstrProcessor", "MgpstrTokenizer"], + "models.mistral": ["MISTRAL_PRETRAINED_CONFIG_ARCHIVE_MAP", "MistralConfig"], "models.mluke": [], "models.mobilebert": ["MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MobileBertConfig", "MobileBertTokenizer"], "models.mobilenet_v1": ["MOBILENET_V1_PRETRAINED_CONFIG_ARCHIVE_MAP", "MobileNetV1Config"], @@ -2222,6 +2223,9 @@ else: "MgpstrPreTrainedModel", ] ) + _import_structure["models.mistral"].extend( + ["MistralForCausalLM", "MistralForSequenceClassification", "MistralModel", "MistralPreTrainedModel"] + ) _import_structure["models.mobilebert"].extend( [ "MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST", @@ -4551,6 +4555,7 @@ if TYPE_CHECKING: from .models.mega import MEGA_PRETRAINED_CONFIG_ARCHIVE_MAP, MegaConfig from .models.megatron_bert import MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MegatronBertConfig from .models.mgp_str import MGP_STR_PRETRAINED_CONFIG_ARCHIVE_MAP, MgpstrConfig, MgpstrProcessor, MgpstrTokenizer + from .models.mistral import MISTRAL_PRETRAINED_CONFIG_ARCHIVE_MAP, MistralConfig from .models.mobilebert import MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MobileBertConfig, MobileBertTokenizer from .models.mobilenet_v1 import MOBILENET_V1_PRETRAINED_CONFIG_ARCHIVE_MAP, MobileNetV1Config from .models.mobilenet_v2 import MOBILENET_V2_PRETRAINED_CONFIG_ARCHIVE_MAP, MobileNetV2Config @@ -6069,6 +6074,12 @@ if TYPE_CHECKING: MgpstrModel, MgpstrPreTrainedModel, ) + from .models.mistral import ( + MistralForCausalLM, + MistralForSequenceClassification, + MistralModel, + MistralPreTrainedModel, + ) from .models.mobilebert import ( MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST, MobileBertForMaskedLM, diff --git a/src/transformers/models/__init__.py b/src/transformers/models/__init__.py index 368d4db85d..e98f672f8d 100644 --- a/src/transformers/models/__init__.py +++ b/src/transformers/models/__init__.py @@ -131,6 +131,7 @@ from . import ( megatron_bert, megatron_gpt2, mgp_str, + mistral, mluke, mobilebert, mobilenet_v1, diff --git a/src/transformers/models/auto/configuration_auto.py b/src/transformers/models/auto/configuration_auto.py index 80a0c6bf4a..97f4ad3354 100755 --- a/src/transformers/models/auto/configuration_auto.py +++ b/src/transformers/models/auto/configuration_auto.py @@ -138,6 +138,7 @@ CONFIG_MAPPING_NAMES = OrderedDict( ("mega", "MegaConfig"), ("megatron-bert", "MegatronBertConfig"), ("mgp-str", "MgpstrConfig"), + ("mistral", "MistralConfig"), ("mobilebert", "MobileBertConfig"), ("mobilenet_v1", "MobileNetV1Config"), ("mobilenet_v2", "MobileNetV2Config"), @@ -346,6 +347,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict( ("mega", "MEGA_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("megatron-bert", "MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("mgp-str", "MGP_STR_PRETRAINED_CONFIG_ARCHIVE_MAP"), + ("mistral", "MISTRAL_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("mobilenet_v1", "MOBILENET_V1_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("mobilenet_v2", "MOBILENET_V2_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("mobilevit", "MOBILEVIT_PRETRAINED_CONFIG_ARCHIVE_MAP"), @@ -564,6 +566,7 @@ MODEL_NAMES_MAPPING = OrderedDict( ("megatron-bert", "Megatron-BERT"), ("megatron_gpt2", "Megatron-GPT2"), ("mgp-str", "MGP-STR"), + ("mistral", "Mistral"), ("mluke", "mLUKE"), ("mms", "MMS"), ("mobilebert", "MobileBERT"), diff --git a/src/transformers/models/auto/modeling_auto.py b/src/transformers/models/auto/modeling_auto.py index c21a3d27b4..0ad5994aca 100755 --- a/src/transformers/models/auto/modeling_auto.py +++ b/src/transformers/models/auto/modeling_auto.py @@ -134,6 +134,7 @@ MODEL_MAPPING_NAMES = OrderedDict( ("mega", "MegaModel"), ("megatron-bert", "MegatronBertModel"), ("mgp-str", "MgpstrForSceneTextRecognition"), + ("mistral", "MistralModel"), ("mobilebert", "MobileBertModel"), ("mobilenet_v1", "MobileNetV1Model"), ("mobilenet_v2", "MobileNetV2Model"), @@ -411,6 +412,7 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict( ("mbart", "MBartForCausalLM"), ("mega", "MegaForCausalLM"), ("megatron-bert", "MegatronBertForCausalLM"), + ("mistral", "MistralForCausalLM"), ("mpt", "MptForCausalLM"), ("musicgen", "MusicgenForCausalLM"), ("mvp", "MvpForCausalLM"), @@ -732,6 +734,7 @@ MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES = OrderedDict( ("mbart", "MBartForSequenceClassification"), ("mega", "MegaForSequenceClassification"), ("megatron-bert", "MegatronBertForSequenceClassification"), + ("mistral", "MistralForSequenceClassification"), ("mobilebert", "MobileBertForSequenceClassification"), ("mpnet", "MPNetForSequenceClassification"), ("mpt", "MptForSequenceClassification"), diff --git a/src/transformers/models/auto/tokenization_auto.py b/src/transformers/models/auto/tokenization_auto.py index 0f72f56586..094d3351e8 100644 --- a/src/transformers/models/auto/tokenization_auto.py +++ b/src/transformers/models/auto/tokenization_auto.py @@ -223,6 +223,13 @@ else: ("mega", ("RobertaTokenizer", "RobertaTokenizerFast" if is_tokenizers_available() else None)), ("megatron-bert", ("BertTokenizer", "BertTokenizerFast" if is_tokenizers_available() else None)), ("mgp-str", ("MgpstrTokenizer", None)), + ( + "mistral", + ( + "LlamaTokenizer" if is_sentencepiece_available() else None, + "LlamaTokenizerFast" if is_tokenizers_available() else None, + ), + ), ("mluke", ("MLukeTokenizer" if is_sentencepiece_available() else None, None)), ("mobilebert", ("MobileBertTokenizer", "MobileBertTokenizerFast" if is_tokenizers_available() else None)), ("mpnet", ("MPNetTokenizer", "MPNetTokenizerFast" if is_tokenizers_available() else None)), diff --git a/src/transformers/models/mistral/__init__.py b/src/transformers/models/mistral/__init__.py new file mode 100644 index 0000000000..2f308031dd --- /dev/null +++ b/src/transformers/models/mistral/__init__.py @@ -0,0 +1,62 @@ +# Copyright 2023 Mistral AI and The HuggingFace Inc. team. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +from typing import TYPE_CHECKING + +from ...utils import ( + OptionalDependencyNotAvailable, + _LazyModule, + is_torch_available, +) + + +_import_structure = { + "configuration_mistral": ["MISTRAL_PRETRAINED_CONFIG_ARCHIVE_MAP", "MistralConfig"], +} + + +try: + if not is_torch_available(): + raise OptionalDependencyNotAvailable() +except OptionalDependencyNotAvailable: + pass +else: + _import_structure["modeling_mistral"] = [ + "MistralForCausalLM", + "MistralModel", + "MistralPreTrainedModel", + "MistralForSequenceClassification", + ] + + +if TYPE_CHECKING: + from .configuration_mistral import MISTRAL_PRETRAINED_CONFIG_ARCHIVE_MAP, MistralConfig + + try: + if not is_torch_available(): + raise OptionalDependencyNotAvailable() + except OptionalDependencyNotAvailable: + pass + else: + from .modeling_mistral import ( + MistralForCausalLM, + MistralForSequenceClassification, + MistralModel, + MistralPreTrainedModel, + ) + + +else: + import sys + + sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__) diff --git a/src/transformers/models/mistral/configuration_mistral.py b/src/transformers/models/mistral/configuration_mistral.py new file mode 100644 index 0000000000..b308deb719 --- /dev/null +++ b/src/transformers/models/mistral/configuration_mistral.py @@ -0,0 +1,144 @@ +# coding=utf-8 +# Copyright 2023 Mistral AI and the HuggingFace Inc. team. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +""" Mistral model configuration""" + +from ...configuration_utils import PretrainedConfig +from ...utils import logging + + +logger = logging.get_logger(__name__) + +MISTRAL_PRETRAINED_CONFIG_ARCHIVE_MAP = { + "mistralai/Mistral-7B-v0.1": "https://huggingface.co/mistralai/Mistral-7B-v0.1/resolve/main/config.json", + "mistralai/Mistral-7B-Instruct-v0.1": "https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1/resolve/main/config.json", +} + + +class MistralConfig(PretrainedConfig): + r""" + This is the configuration class to store the configuration of a [`MistralModel`]. It is used to instantiate an + Mistral model according to the specified arguments, defining the model architecture. Instantiating a configuration + with the defaults will yield a similar configuration to that of the Mistral-7B-v0.1 or Mistral-7B-Instruct-v0.1. + + [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) + [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) + + Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the + documentation from [`PretrainedConfig`] for more information. + + + Args: + vocab_size (`int`, *optional*, defaults to 32000): + Vocabulary size of the Mistral model. Defines the number of different tokens that can be represented by the + `inputs_ids` passed when calling [`MistralModel`] + hidden_size (`int`, *optional*, defaults to 4096): + Dimension of the hidden representations. + intermediate_size (`int`, *optional*, defaults to 14336): + Dimension of the MLP representations. + num_hidden_layers (`int`, *optional*, defaults to 32): + Number of hidden layers in the Transformer encoder. + num_attention_heads (`int`, *optional*, defaults to 32): + Number of attention heads for each attention layer in the Transformer encoder. + num_key_value_heads (`int`, *optional*, defaults to 8): + This is the number of key_value heads that should be used to implement Grouped Query Attention. If + `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if + `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When + converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed + by meanpooling all the original heads within that group. For more details checkout [this + paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `8`. + hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): + The non-linear activation function (function or string) in the decoder. + max_position_embeddings (`int`, *optional*, defaults to 4096*32): + The maximum sequence length that this model might ever be used with. Mistral's sliding window attention + allows sequence of up to 4096*32 tokens. + initializer_range (`float`, *optional*, defaults to 0.02): + The standard deviation of the truncated_normal_initializer for initializing all weight matrices. + rms_norm_eps (`float`, *optional*, defaults to 1e-12): + The epsilon used by the rms normalization layers. + use_cache (`bool`, *optional*, defaults to `True`): + Whether or not the model should return the last key/values attentions (not used by all models). Only + relevant if `config.is_decoder=True`. + tie_word_embeddings(`bool`, *optional*, defaults to `False`): + Whether to tie weight embeddings + rope_theta (`float`, *optional*, defaults to 10000.0): + The base period of the RoPE embeddings. + sliding_window (`int`, *optional*, defaults to 4096): + Sliding window attention window size. If not specified, will default to `4096`. + + Example: + + ```python + >>> from transformers import MistralModel, MistralConfig + + >>> # Initializing a Mistral 7B style configuration + >>> configuration = MistralConfig() + + >>> # Initializing a model from the Mistral 7B style configuration + >>> model = MistralModel(configuration) + + >>> # Accessing the model configuration + >>> configuration = model.config + ```""" + + model_type = "mistral" + keys_to_ignore_at_inference = ["past_key_values"] + + def __init__( + self, + vocab_size=32000, + hidden_size=4096, + intermediate_size=14336, + num_hidden_layers=32, + num_attention_heads=32, + num_key_value_heads=8, + hidden_act="silu", + max_position_embeddings=4096 * 32, + initializer_range=0.02, + rms_norm_eps=1e-6, + use_cache=True, + pad_token_id=None, + bos_token_id=1, + eos_token_id=2, + tie_word_embeddings=False, + rope_theta=10000.0, + sliding_window=4096, + **kwargs, + ): + self.vocab_size = vocab_size + self.max_position_embeddings = max_position_embeddings + self.hidden_size = hidden_size + self.intermediate_size = intermediate_size + self.num_hidden_layers = num_hidden_layers + self.num_attention_heads = num_attention_heads + self.sliding_window = sliding_window + + # for backward compatibility + if num_key_value_heads is None: + num_key_value_heads = num_attention_heads + + self.num_key_value_heads = num_key_value_heads + self.hidden_act = hidden_act + self.initializer_range = initializer_range + self.rms_norm_eps = rms_norm_eps + self.use_cache = use_cache + self.rope_theta = rope_theta + + super().__init__( + pad_token_id=pad_token_id, + bos_token_id=bos_token_id, + eos_token_id=eos_token_id, + tie_word_embeddings=tie_word_embeddings, + **kwargs, + ) diff --git a/src/transformers/models/mistral/convert_mistral_weights_to_hf.py b/src/transformers/models/mistral/convert_mistral_weights_to_hf.py new file mode 100644 index 0000000000..6e1ffbcc04 --- /dev/null +++ b/src/transformers/models/mistral/convert_mistral_weights_to_hf.py @@ -0,0 +1,276 @@ +# Copyright 2023 Mistral AI and The HuggingFace Inc. team. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import argparse +import gc +import json +import os +import shutil +import warnings + +import torch + +from transformers import ( + LlamaTokenizer, + MistralConfig, + MistralForCausalLM, +) + + +try: + from transformers import LlamaTokenizerFast + + tokenizer_class = LlamaTokenizerFast +except ImportError as e: + warnings.warn(e) + warnings.warn( + "The converted tokenizer will be the `slow` tokenizer. To use the fast, update your `tokenizers` library and re-run the tokenizer conversion" + ) + tokenizer_class = LlamaTokenizer + +""" +Sample usage: + +``` +python src/transformers/models/mistral/convert_mistral_weights_to_hf.py \ + --input_dir /path/to/downloaded/mistral/weights --model_size 7B --output_dir /output/path +``` + +Thereafter, models can be loaded via: + +```py +from transformers import MistralForCausalLM, LlamaTokenizer + +model = MistralForCausalLM.from_pretrained("/output/path") +tokenizer = LlamaTokenizer.from_pretrained("/output/path") +``` + +Important note: you need to be able to host the whole model in RAM to execute this script (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). +""" + +NUM_SHARDS = {"7B": 1} + + +def compute_intermediate_size(n, ffn_dim_multiplier=1, multiple_of=256): + return multiple_of * ((int(ffn_dim_multiplier * int(8 * n / 3)) + multiple_of - 1) // multiple_of) + + +def read_json(path): + with open(path, "r") as f: + return json.load(f) + + +def write_json(text, path): + with open(path, "w") as f: + json.dump(text, f) + + +def write_model(model_path, input_base_path, model_size, tokenizer_path=None, safe_serialization=True): + # for backward compatibility, before you needed the repo to be called `my_repo/model_size` + if not os.path.isfile(os.path.join(input_base_path, "params.json")): + input_base_path = os.path.join(input_base_path, model_size) + + os.makedirs(model_path, exist_ok=True) + tmp_model_path = os.path.join(model_path, "tmp") + os.makedirs(tmp_model_path, exist_ok=True) + + params = read_json(os.path.join(input_base_path, "params.json")) + num_shards = NUM_SHARDS[model_size] + + # For some reason this is a string in the params.json + sliding_window = int(params["ragged_attention"]) + n_layers = params["n_layers"] + n_heads = params["n_heads"] + n_heads_per_shard = n_heads // num_shards + dim = params["dim"] + dims_per_head = dim // n_heads + base = params.get("rope_theta", 10000.0) + inv_freq = 1.0 / (base ** (torch.arange(0, dims_per_head, 2).float() / dims_per_head)) + max_position_embeddings = 4096 * 8 + + if tokenizer_path is not None: + tokenizer = tokenizer_class(tokenizer_path) + tokenizer.save_pretrained(model_path) + vocab_size = tokenizer.vocab_size if tokenizer_path is not None else 32000 + + if "n_kv_heads" in params: + num_key_value_heads = params["n_kv_heads"] # for GQA / MQA + num_local_key_value_heads = num_key_value_heads // num_shards + key_value_dim = dims_per_head * num_local_key_value_heads + else: # compatibility with other checkpoints + num_key_value_heads = n_heads + num_local_key_value_heads = n_heads_per_shard + key_value_dim = dim + + # permute for sliced rotary + def permute(w, n_heads=n_heads, dim1=dim, dim2=dim): + return w.view(n_heads, dim1 // n_heads // 2, 2, dim2).transpose(1, 2).reshape(dim1, dim2) + + print(f"Fetching all parameters from the checkpoint at {input_base_path}.") + # Load weights + loaded = [ + torch.load(os.path.join(input_base_path, f"consolidated.{i:02d}.pth"), map_location="cpu") + for i in range(num_shards) + ] + param_count = 0 + index_dict = {"weight_map": {}} + for layer_i in range(n_layers): + filename = f"pytorch_model-{layer_i + 1}-of-{n_layers + 1}.bin" + + # Sharded + # Note that attention.w{q,k,v,o}, feed_fordward.w[1,2,3], attention_norm.weight and ffn_norm.weight share + # the same storage object, saving attention_norm and ffn_norm will save other weights too, which is + # redundant as other weights will be stitched from multiple shards. To avoid that, they are cloned. + + state_dict = { + f"model.layers.{layer_i}.input_layernorm.weight": loaded[0][ + f"layers.{layer_i}.attention_norm.weight" + ].clone(), + f"model.layers.{layer_i}.post_attention_layernorm.weight": loaded[0][ + f"layers.{layer_i}.ffn_norm.weight" + ].clone(), + } + state_dict[f"model.layers.{layer_i}.self_attn.q_proj.weight"] = permute( + torch.cat( + [ + loaded[i][f"layers.{layer_i}.attention.wq.weight"].view(n_heads_per_shard, dims_per_head, dim) + for i in range(num_shards) + ], + dim=0, + ).reshape(dim, dim) + ) + state_dict[f"model.layers.{layer_i}.self_attn.k_proj.weight"] = permute( + torch.cat( + [ + loaded[i][f"layers.{layer_i}.attention.wk.weight"].view( + num_local_key_value_heads, dims_per_head, dim + ) + for i in range(num_shards) + ], + dim=0, + ).reshape(key_value_dim, dim), + num_key_value_heads, + key_value_dim, + dim, + ) + state_dict[f"model.layers.{layer_i}.self_attn.v_proj.weight"] = torch.cat( + [ + loaded[i][f"layers.{layer_i}.attention.wv.weight"].view(num_local_key_value_heads, dims_per_head, dim) + for i in range(num_shards) + ], + dim=0, + ).reshape(key_value_dim, dim) + + state_dict[f"model.layers.{layer_i}.self_attn.o_proj.weight"] = torch.cat( + [loaded[i][f"layers.{layer_i}.attention.wo.weight"] for i in range(num_shards)], dim=1 + ) + state_dict[f"model.layers.{layer_i}.mlp.gate_proj.weight"] = torch.cat( + [loaded[i][f"layers.{layer_i}.feed_forward.w1.weight"] for i in range(num_shards)], dim=0 + ) + state_dict[f"model.layers.{layer_i}.mlp.down_proj.weight"] = torch.cat( + [loaded[i][f"layers.{layer_i}.feed_forward.w2.weight"] for i in range(num_shards)], dim=1 + ) + state_dict[f"model.layers.{layer_i}.mlp.up_proj.weight"] = torch.cat( + [loaded[i][f"layers.{layer_i}.feed_forward.w3.weight"] for i in range(num_shards)], dim=0 + ) + + state_dict[f"model.layers.{layer_i}.self_attn.rotary_emb.inv_freq"] = inv_freq + for k, v in state_dict.items(): + index_dict["weight_map"][k] = filename + param_count += v.numel() + torch.save(state_dict, os.path.join(tmp_model_path, filename)) + + filename = f"pytorch_model-{n_layers + 1}-of-{n_layers + 1}.bin" + state_dict = { + "model.norm.weight": loaded[0]["norm.weight"], + "model.embed_tokens.weight": torch.cat([loaded[i]["tok_embeddings.weight"] for i in range(num_shards)], dim=1), + "lm_head.weight": torch.cat([loaded[i]["output.weight"] for i in range(num_shards)], dim=0), + } + + for k, v in state_dict.items(): + index_dict["weight_map"][k] = filename + param_count += v.numel() + torch.save(state_dict, os.path.join(tmp_model_path, filename)) + + # Write configs + index_dict["metadata"] = {"total_size": param_count * 2} + write_json(index_dict, os.path.join(tmp_model_path, "pytorch_model.bin.index.json")) + config = MistralConfig( + hidden_size=dim, + intermediate_size=params["hidden_dim"], + num_attention_heads=params["n_heads"], + num_hidden_layers=params["n_layers"], + rms_norm_eps=params["norm_eps"], + num_key_value_heads=num_key_value_heads, + vocab_size=vocab_size, + rope_theta=base, + max_position_embeddings=max_position_embeddings, + sliding_window=sliding_window, + ) + config.save_pretrained(tmp_model_path) + + # Make space so we can load the model properly now. + del state_dict + del loaded + gc.collect() + + print("Loading the checkpoint in a Mistral model.") + model = MistralForCausalLM.from_pretrained(tmp_model_path, torch_dtype=torch.bfloat16, low_cpu_mem_usage=True) + # Avoid saving this as part of the config. + del model.config._name_or_path + model.config.torch_dtype = torch.float16 + print("Saving in the Transformers format.") + model.save_pretrained(model_path, safe_serialization=safe_serialization) + shutil.rmtree(tmp_model_path) + + +def write_tokenizer(tokenizer_path, input_tokenizer_path): + # Initialize the tokenizer based on the `spm` model + print(f"Saving a {tokenizer_class.__name__} to {tokenizer_path}.") + tokenizer = tokenizer_class(input_tokenizer_path) + tokenizer.save_pretrained(tokenizer_path) + + +def main(): + parser = argparse.ArgumentParser() + parser.add_argument( + "--input_dir", + help="Location of Mistral weights, which contains tokenizer.model and model folders", + ) + parser.add_argument( + "--model_size", + choices=["7B", "tokenizer_only"], + help="'f' models correspond to the finetuned versions, and are specific to the Mistral2 official release. For more details on Mistral2, checkout the original repo: https://huggingface.co/meta-mistral", + ) + parser.add_argument( + "--output_dir", + help="Location to write HF model and tokenizer", + ) + parser.add_argument("--safe_serialization", type=bool, help="Whether or not to save using `safetensors`.") + args = parser.parse_args() + spm_path = os.path.join(args.input_dir, "tokenizer.model") + if args.model_size != "tokenizer_only": + write_model( + model_path=args.output_dir, + input_base_path=args.input_dir, + model_size=args.model_size, + safe_serialization=args.safe_serialization, + tokenizer_path=spm_path, + ) + else: + write_tokenizer(args.output_dir, spm_path) + + +if __name__ == "__main__": + main() diff --git a/src/transformers/models/mistral/modeling_mistral.py b/src/transformers/models/mistral/modeling_mistral.py new file mode 100644 index 0000000000..62cfca2946 --- /dev/null +++ b/src/transformers/models/mistral/modeling_mistral.py @@ -0,0 +1,934 @@ +# coding=utf-8 +# Copyright 2023 Mistral AI and the HuggingFace Inc. team. All rights reserved. +# +# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX +# and OPT implementations in this library. It has been modified from its +# original forms to accommodate minor architectural differences compared +# to GPT-NeoX and OPT used by the Meta AI team that trained the model. +# +# 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. +""" PyTorch Mistral model.""" +import math +from typing import List, Optional, Tuple, Union + +import torch +import torch.utils.checkpoint +from torch import nn +from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss + +from ...activations import ACT2FN +from ...modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast +from ...modeling_utils import PreTrainedModel +from ...utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings +from .configuration_mistral import MistralConfig + + +logger = logging.get_logger(__name__) + +_CONFIG_FOR_DOC = "MistralConfig" + + +def _make_sliding_window_causal_mask( + input_ids_shape: torch.Size, + dtype: torch.dtype, + device: torch.device, + past_key_values_length: int = 0, + sliding_window: int = 4096, +): + """ + Make causal mask used for sliding window attention + """ + bsz, tgt_len = input_ids_shape + + tensor = torch.full( + (tgt_len, tgt_len), + fill_value=1, + device=device, + ) + mask = torch.tril(tensor, diagonal=0) + # make the mask banded to account for sliding window + mask = torch.triu(mask, diagonal=-sliding_window) + mask = torch.log(mask).to(dtype) + + if past_key_values_length > 0: + mask = torch.cat([torch.zeros(tgt_len, past_key_values_length, dtype=dtype, device=device), mask], dim=-1) + return mask[None, None, :, :].expand(bsz, 1, tgt_len, tgt_len + past_key_values_length) + + +# Copied from transformers.models.bart.modeling_bart._expand_mask +def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None): + """ + Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`. + """ + bsz, src_len = mask.size() + tgt_len = tgt_len if tgt_len is not None else src_len + + expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype) + + inverted_mask = 1.0 - expanded_mask + + return inverted_mask.masked_fill(inverted_mask.to(torch.bool), torch.finfo(dtype).min) + + +# Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->Mistral +class MistralRMSNorm(nn.Module): + def __init__(self, hidden_size, eps=1e-6): + """ + MistralRMSNorm is equivalent to T5LayerNorm + """ + super().__init__() + self.weight = nn.Parameter(torch.ones(hidden_size)) + self.variance_epsilon = eps + + def forward(self, hidden_states): + input_dtype = hidden_states.dtype + hidden_states = hidden_states.to(torch.float32) + variance = hidden_states.pow(2).mean(-1, keepdim=True) + hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon) + return self.weight * hidden_states.to(input_dtype) + + +# Copied from transformers.models.llama.modeling_llama.LlamaRotaryEmbedding with Llama->Mistral +class MistralRotaryEmbedding(nn.Module): + def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None): + super().__init__() + + self.dim = dim + self.max_position_embeddings = max_position_embeddings + self.base = base + inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim)) + self.register_buffer("inv_freq", inv_freq, persistent=False) + + # Build here to make `torch.jit.trace` work. + self._set_cos_sin_cache( + seq_len=max_position_embeddings, device=self.inv_freq.device, dtype=torch.get_default_dtype() + ) + + def _set_cos_sin_cache(self, seq_len, device, dtype): + self.max_seq_len_cached = seq_len + t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype) + + freqs = torch.einsum("i,j->ij", t, self.inv_freq) + # Different from paper, but it uses a different permutation in order to obtain the same calculation + emb = torch.cat((freqs, freqs), dim=-1) + self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False) + self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False) + + def forward(self, x, seq_len=None): + # x: [bs, num_attention_heads, seq_len, head_size] + if seq_len > self.max_seq_len_cached: + self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype) + + return ( + self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype), + self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype), + ) + + +# Copied from transformers.models.llama.modeling_llama.rotate_half +def rotate_half(x): + """Rotates half the hidden dims of the input.""" + x1 = x[..., : x.shape[-1] // 2] + x2 = x[..., x.shape[-1] // 2 :] + return torch.cat((-x2, x1), dim=-1) + + +# Copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb +def apply_rotary_pos_emb(q, k, cos, sin, position_ids): + # The first two dimensions of cos and sin are always 1, so we can `squeeze` them. + cos = cos.squeeze(1).squeeze(0) # [seq_len, dim] + sin = sin.squeeze(1).squeeze(0) # [seq_len, dim] + cos = cos[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim] + sin = sin[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim] + q_embed = (q * cos) + (rotate_half(q) * sin) + k_embed = (k * cos) + (rotate_half(k) * sin) + return q_embed, k_embed + + +class MistralMLP(nn.Module): + def __init__(self, config): + super().__init__() + self.config = config + self.hidden_size = config.hidden_size + self.intermediate_size = config.intermediate_size + self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False) + self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False) + self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False) + self.act_fn = ACT2FN[config.hidden_act] + + def forward(self, x): + return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x)) + + +def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor: + """ + This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch, + num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim) + """ + batch, num_key_value_heads, slen, head_dim = hidden_states.shape + if n_rep == 1: + return hidden_states + hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim) + return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim) + + +class MistralAttention(nn.Module): + """ + Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer + and "Generating Long Sequences with Sparse Transformers". + """ + + def __init__(self, config: MistralConfig): + super().__init__() + self.config = config + self.hidden_size = config.hidden_size + self.num_heads = config.num_attention_heads + self.head_dim = self.hidden_size // self.num_heads + self.num_key_value_heads = config.num_key_value_heads + self.num_key_value_groups = self.num_heads // self.num_key_value_heads + self.max_position_embeddings = config.max_position_embeddings + self.rope_theta = config.rope_theta + + if (self.head_dim * self.num_heads) != self.hidden_size: + raise ValueError( + f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}" + f" and `num_heads`: {self.num_heads})." + ) + self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False) + self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False) + self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False) + self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False) + + self.rotary_emb = MistralRotaryEmbedding( + self.head_dim, + max_position_embeddings=self.max_position_embeddings, + base=self.rope_theta, + ) + + def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int): + return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous() + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_value: Optional[Tuple[torch.Tensor]] = None, + output_attentions: bool = False, + use_cache: bool = False, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: + bsz, q_len, _ = hidden_states.size() + + query_states = self.q_proj(hidden_states) + key_states = self.k_proj(hidden_states) + value_states = self.v_proj(hidden_states) + + query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) + key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + + kv_seq_len = key_states.shape[-2] + if past_key_value is not None: + kv_seq_len += past_key_value[0].shape[-2] + cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len) + query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids) + + if past_key_value is not None: + # reuse k, v, self_attention + key_states = torch.cat([past_key_value[0], key_states], dim=2) + value_states = torch.cat([past_key_value[1], value_states], dim=2) + + past_key_value = (key_states, value_states) if use_cache else None + + # repeat k/v heads if n_kv_heads < n_heads + key_states = repeat_kv(key_states, self.num_key_value_groups) + value_states = repeat_kv(value_states, self.num_key_value_groups) + + attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim) + + if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len): + raise ValueError( + f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is" + f" {attn_weights.size()}" + ) + + if attention_mask is not None: + if attention_mask.size() != (bsz, 1, q_len, kv_seq_len): + raise ValueError( + f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}" + ) + + attn_weights = attn_weights + attention_mask + + # upcast attention to fp32 + attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype) + attn_output = torch.matmul(attn_weights, value_states) + + if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim): + raise ValueError( + f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is" + f" {attn_output.size()}" + ) + + attn_output = attn_output.transpose(1, 2).contiguous() + attn_output = attn_output.reshape(bsz, q_len, self.hidden_size) + + attn_output = self.o_proj(attn_output) + + if not output_attentions: + attn_weights = None + + return attn_output, attn_weights, past_key_value + + +class MistralDecoderLayer(nn.Module): + def __init__(self, config: MistralConfig): + super().__init__() + self.hidden_size = config.hidden_size + self.self_attn = MistralAttention(config=config) + self.mlp = MistralMLP(config) + self.input_layernorm = MistralRMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.post_attention_layernorm = MistralRMSNorm(config.hidden_size, eps=config.rms_norm_eps) + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_value: Optional[Tuple[torch.Tensor]] = None, + output_attentions: Optional[bool] = False, + use_cache: Optional[bool] = False, + ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]: + """ + Args: + hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` + attention_mask (`torch.FloatTensor`, *optional*): attention mask of size + `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values. + output_attentions (`bool`, *optional*): + Whether or not to return the attentions tensors of all attention layers. See `attentions` under + returned tensors for more detail. + use_cache (`bool`, *optional*): + If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding + (see `past_key_values`). + past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states + """ + + residual = hidden_states + + hidden_states = self.input_layernorm(hidden_states) + + # Self Attention + hidden_states, self_attn_weights, present_key_value = self.self_attn( + hidden_states=hidden_states, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_value=past_key_value, + output_attentions=output_attentions, + use_cache=use_cache, + ) + hidden_states = residual + hidden_states + + # Fully Connected + residual = hidden_states + hidden_states = self.post_attention_layernorm(hidden_states) + hidden_states = self.mlp(hidden_states) + hidden_states = residual + hidden_states + + outputs = (hidden_states,) + + if output_attentions: + outputs += (self_attn_weights,) + + if use_cache: + outputs += (present_key_value,) + + return outputs + + +MISTRAL_START_DOCSTRING = r""" + This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the + library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads + etc.) + + This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass. + Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage + and behavior. + + Parameters: + config ([`MistralConfig`]): + Model configuration class with all the parameters of the model. Initializing with a config file does not + load the weights associated with the model, only the configuration. Check out the + [`~PreTrainedModel.from_pretrained`] method to load the model weights. +""" + + +@add_start_docstrings( + "The bare Mistral Model outputting raw hidden-states without any specific head on top.", + MISTRAL_START_DOCSTRING, +) +class MistralPreTrainedModel(PreTrainedModel): + config_class = MistralConfig + base_model_prefix = "model" + supports_gradient_checkpointing = True + _no_split_modules = ["MistralDecoderLayer"] + _skip_keys_device_placement = "past_key_values" + + def _init_weights(self, module): + std = self.config.initializer_range + if isinstance(module, nn.Linear): + module.weight.data.normal_(mean=0.0, std=std) + if module.bias is not None: + module.bias.data.zero_() + elif isinstance(module, nn.Embedding): + module.weight.data.normal_(mean=0.0, std=std) + if module.padding_idx is not None: + module.weight.data[module.padding_idx].zero_() + + def _set_gradient_checkpointing(self, module, value=False): + if isinstance(module, MistralModel): + module.gradient_checkpointing = value + + +MISTRAL_INPUTS_DOCSTRING = r""" + Args: + input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): + Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide + it. + + Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and + [`PreTrainedTokenizer.__call__`] for details. + + [What are input IDs?](../glossary#input-ids) + attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*): + Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`: + + - 1 for tokens that are **not masked**, + - 0 for tokens that are **masked**. + + [What are attention masks?](../glossary#attention-mask) + + Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and + [`PreTrainedTokenizer.__call__`] for details. + + If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see + `past_key_values`). + + If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`] + and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more + information on the default strategy. + + - 1 indicates the head is **not masked**, + - 0 indicates the head is **masked**. + position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0, + config.n_positions - 1]`. + + [What are position IDs?](../glossary#position-ids) + past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`): + Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape + `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape + `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`. + + Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention + blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. + + If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that + don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all + `decoder_input_ids` of shape `(batch_size, sequence_length)`. + inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*): + Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This + is useful if you want more control over how to convert `input_ids` indices into associated vectors than the + model's internal embedding lookup matrix. + use_cache (`bool`, *optional*): + If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see + `past_key_values`). + output_attentions (`bool`, *optional*): + Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned + tensors for more detail. + output_hidden_states (`bool`, *optional*): + Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for + more detail. + return_dict (`bool`, *optional*): + Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. +""" + + +@add_start_docstrings( + "The bare Mistral Model outputting raw hidden-states without any specific head on top.", + MISTRAL_START_DOCSTRING, +) +class MistralModel(MistralPreTrainedModel): + """ + Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`MistralDecoderLayer`] + + Args: + config: MistralConfig + """ + + def __init__(self, config: MistralConfig): + super().__init__(config) + self.padding_idx = config.pad_token_id + self.vocab_size = config.vocab_size + + self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx) + self.layers = nn.ModuleList([MistralDecoderLayer(config) for _ in range(config.num_hidden_layers)]) + self.norm = MistralRMSNorm(config.hidden_size, eps=config.rms_norm_eps) + + self.gradient_checkpointing = False + # Initialize weights and apply final processing + self.post_init() + + def get_input_embeddings(self): + return self.embed_tokens + + def set_input_embeddings(self, value): + self.embed_tokens = value + + def _prepare_decoder_attention_mask( + self, attention_mask, input_shape, inputs_embeds, past_key_values_length, sliding_window + ): + # create causal mask + # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len] + combined_attention_mask = None + if input_shape[-1] > 1: + combined_attention_mask = _make_sliding_window_causal_mask( + input_shape, + inputs_embeds.dtype, + device=inputs_embeds.device, + past_key_values_length=past_key_values_length, + sliding_window=sliding_window, + ) + + if attention_mask is not None: + # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len] + expanded_attn_mask = _expand_mask(attention_mask, inputs_embeds.dtype, tgt_len=input_shape[-1]).to( + inputs_embeds.device + ) + combined_attention_mask = ( + expanded_attn_mask if combined_attention_mask is None else expanded_attn_mask + combined_attention_mask + ) + + return combined_attention_mask + + @add_start_docstrings_to_model_forward(MISTRAL_INPUTS_DOCSTRING) + def forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[List[torch.FloatTensor]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple, BaseModelOutputWithPast]: + output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + use_cache = use_cache if use_cache is not None else self.config.use_cache + + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + # retrieve input_ids and inputs_embeds + if input_ids is not None and inputs_embeds is not None: + raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time") + elif input_ids is not None: + batch_size, seq_length = input_ids.shape + elif inputs_embeds is not None: + batch_size, seq_length, _ = inputs_embeds.shape + else: + raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds") + + seq_length_with_past = seq_length + past_key_values_length = 0 + + if past_key_values is not None: + past_key_values_length = past_key_values[0][0].shape[2] + seq_length_with_past = seq_length_with_past + past_key_values_length + + if position_ids is None: + device = input_ids.device if input_ids is not None else inputs_embeds.device + position_ids = torch.arange( + past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device + ) + position_ids = position_ids.unsqueeze(0).view(-1, seq_length) + else: + position_ids = position_ids.view(-1, seq_length).long() + + if inputs_embeds is None: + inputs_embeds = self.embed_tokens(input_ids) + # embed positions + if attention_mask is None: + attention_mask = torch.ones( + (batch_size, seq_length_with_past), dtype=torch.bool, device=inputs_embeds.device + ) + attention_mask = self._prepare_decoder_attention_mask( + attention_mask, + (batch_size, seq_length), + inputs_embeds, + past_key_values_length, + sliding_window=self.config.sliding_window, + ) + + hidden_states = inputs_embeds + + if self.gradient_checkpointing and self.training: + if use_cache: + logger.warning_once( + "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..." + ) + use_cache = False + + # decoder layers + all_hidden_states = () if output_hidden_states else None + all_self_attns = () if output_attentions else None + next_decoder_cache = () if use_cache else None + + for idx, decoder_layer in enumerate(self.layers): + if output_hidden_states: + all_hidden_states += (hidden_states,) + + past_key_value = past_key_values[idx] if past_key_values is not None else None + + if self.gradient_checkpointing and self.training: + + def create_custom_forward(module): + def custom_forward(*inputs): + # None for past_key_value + return module(*inputs, past_key_value, output_attentions) + + return custom_forward + + layer_outputs = torch.utils.checkpoint.checkpoint( + create_custom_forward(decoder_layer), + hidden_states, + attention_mask, + position_ids, + ) + else: + layer_outputs = decoder_layer( + hidden_states, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_value=past_key_value, + output_attentions=output_attentions, + use_cache=use_cache, + ) + + hidden_states = layer_outputs[0] + + if use_cache: + next_decoder_cache += (layer_outputs[2 if output_attentions else 1],) + + if output_attentions: + all_self_attns += (layer_outputs[1],) + + hidden_states = self.norm(hidden_states) + + # add hidden states from the last decoder layer + if output_hidden_states: + all_hidden_states += (hidden_states,) + + next_cache = next_decoder_cache if use_cache else None + if not return_dict: + return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None) + return BaseModelOutputWithPast( + last_hidden_state=hidden_states, + past_key_values=next_cache, + hidden_states=all_hidden_states, + attentions=all_self_attns, + ) + + +class MistralForCausalLM(MistralPreTrainedModel): + _tied_weights_keys = ["lm_head.weight"] + + def __init__(self, config): + super().__init__(config) + self.model = MistralModel(config) + self.vocab_size = config.vocab_size + self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False) + + # Initialize weights and apply final processing + self.post_init() + + def get_input_embeddings(self): + return self.model.embed_tokens + + def set_input_embeddings(self, value): + self.model.embed_tokens = value + + def get_output_embeddings(self): + return self.lm_head + + def set_output_embeddings(self, new_embeddings): + self.lm_head = new_embeddings + + def set_decoder(self, decoder): + self.model = decoder + + def get_decoder(self): + return self.model + + @add_start_docstrings_to_model_forward(MISTRAL_INPUTS_DOCSTRING) + @replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC) + def forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[List[torch.FloatTensor]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple, CausalLMOutputWithPast]: + r""" + Args: + labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Labels for computing the masked language modeling loss. Indices should either be in `[0, ..., + config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored + (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. + + Returns: + + Example: + + ```python + >>> from transformers import AutoTokenizer, MistralForCausalLM + + >>> model = MistralForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS) + >>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER) + + >>> prompt = "Hey, are you conscious? Can you talk to me?" + >>> inputs = tokenizer(prompt, return_tensors="pt") + + >>> # Generate + >>> generate_ids = model.generate(inputs.input_ids, max_length=30) + >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0] + "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you." + ```""" + + output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn) + outputs = self.model( + input_ids=input_ids, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_values=past_key_values, + inputs_embeds=inputs_embeds, + use_cache=use_cache, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + hidden_states = outputs[0] + logits = self.lm_head(hidden_states) + logits = logits.float() + + loss = None + if labels is not None: + # Shift so that tokens < n predict n + shift_logits = logits[..., :-1, :].contiguous() + shift_labels = labels[..., 1:].contiguous() + # Flatten the tokens + loss_fct = CrossEntropyLoss() + shift_logits = shift_logits.view(-1, self.config.vocab_size) + shift_labels = shift_labels.view(-1) + # Enable model parallelism + shift_labels = shift_labels.to(shift_logits.device) + loss = loss_fct(shift_logits, shift_labels) + + if not return_dict: + output = (logits,) + outputs[1:] + return (loss,) + output if loss is not None else output + + return CausalLMOutputWithPast( + loss=loss, + logits=logits, + past_key_values=outputs.past_key_values, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + def prepare_inputs_for_generation( + self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs + ): + if past_key_values: + input_ids = input_ids[:, -1:] + + position_ids = kwargs.get("position_ids", None) + if attention_mask is not None and position_ids is None: + # create position_ids on the fly for batch generation + position_ids = attention_mask.long().cumsum(-1) - 1 + position_ids.masked_fill_(attention_mask == 0, 1) + if past_key_values: + position_ids = position_ids[:, -1].unsqueeze(-1) + + # if `inputs_embeds` are passed, we only want to use them in the 1st generation step + if inputs_embeds is not None and past_key_values is None: + model_inputs = {"inputs_embeds": inputs_embeds} + else: + model_inputs = {"input_ids": input_ids} + + model_inputs.update( + { + "position_ids": position_ids, + "past_key_values": past_key_values, + "use_cache": kwargs.get("use_cache"), + "attention_mask": attention_mask, + } + ) + return model_inputs + + @staticmethod + def _reorder_cache(past_key_values, beam_idx): + reordered_past = () + for layer_past in past_key_values: + reordered_past += ( + tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past), + ) + return reordered_past + + +@add_start_docstrings( + """ + The Mistral Model transformer with a sequence classification head on top (linear layer). + + [`MistralForSequenceClassification`] uses the last token in order to do the classification, as other causal models + (e.g. GPT-2) do. + + Since it does classification on the last token, it requires to know the position of the last token. If a + `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If + no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the + padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in + each row of the batch). + """, + MISTRAL_START_DOCSTRING, +) +# Copied from transformers.models.llama.modeling_llama.LlamaForSequenceClassification with Llama->Mistral, LLAMA->MISTRAL +class MistralForSequenceClassification(MistralPreTrainedModel): + def __init__(self, config): + super().__init__(config) + self.num_labels = config.num_labels + self.model = MistralModel(config) + self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False) + + # Initialize weights and apply final processing + self.post_init() + + def get_input_embeddings(self): + return self.model.embed_tokens + + def set_input_embeddings(self, value): + self.model.embed_tokens = value + + @add_start_docstrings_to_model_forward(MISTRAL_INPUTS_DOCSTRING) + def forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[List[torch.FloatTensor]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple, SequenceClassifierOutputWithPast]: + r""" + labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): + Labels for computing the sequence classification/regression loss. Indices should be in `[0, ..., + config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If + `config.num_labels > 1` a classification loss is computed (Cross-Entropy). + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + transformer_outputs = self.model( + input_ids, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_values=past_key_values, + inputs_embeds=inputs_embeds, + use_cache=use_cache, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + hidden_states = transformer_outputs[0] + logits = self.score(hidden_states) + + if input_ids is not None: + batch_size = input_ids.shape[0] + else: + batch_size = inputs_embeds.shape[0] + + if self.config.pad_token_id is None and batch_size != 1: + raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.") + if self.config.pad_token_id is None: + sequence_lengths = -1 + else: + if input_ids is not None: + sequence_lengths = (torch.eq(input_ids, self.config.pad_token_id).long().argmax(-1) - 1).to( + logits.device + ) + else: + sequence_lengths = -1 + + pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths] + + loss = None + if labels is not None: + labels = labels.to(logits.device) + if self.config.problem_type is None: + if self.num_labels == 1: + self.config.problem_type = "regression" + elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int): + self.config.problem_type = "single_label_classification" + else: + self.config.problem_type = "multi_label_classification" + + if self.config.problem_type == "regression": + loss_fct = MSELoss() + if self.num_labels == 1: + loss = loss_fct(pooled_logits.squeeze(), labels.squeeze()) + else: + loss = loss_fct(pooled_logits, labels) + elif self.config.problem_type == "single_label_classification": + loss_fct = CrossEntropyLoss() + loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1)) + elif self.config.problem_type == "multi_label_classification": + loss_fct = BCEWithLogitsLoss() + loss = loss_fct(pooled_logits, labels) + if not return_dict: + output = (pooled_logits,) + transformer_outputs[1:] + return ((loss,) + output) if loss is not None else output + + return SequenceClassifierOutputWithPast( + loss=loss, + logits=pooled_logits, + past_key_values=transformer_outputs.past_key_values, + hidden_states=transformer_outputs.hidden_states, + attentions=transformer_outputs.attentions, + ) diff --git a/src/transformers/utils/dummy_pt_objects.py b/src/transformers/utils/dummy_pt_objects.py index f30b04e0e7..aad1018e19 100644 --- a/src/transformers/utils/dummy_pt_objects.py +++ b/src/transformers/utils/dummy_pt_objects.py @@ -5051,6 +5051,34 @@ class MgpstrPreTrainedModel(metaclass=DummyObject): requires_backends(self, ["torch"]) +class MistralForCausalLM(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class MistralForSequenceClassification(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class MistralModel(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class MistralPreTrainedModel(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST = None diff --git a/tests/models/mistral/__init__.py b/tests/models/mistral/__init__.py new file mode 100644 index 0000000000..e69de29bb2 diff --git a/tests/models/mistral/test_modeling_mistral.py b/tests/models/mistral/test_modeling_mistral.py new file mode 100644 index 0000000000..df3bcf9d67 --- /dev/null +++ b/tests/models/mistral/test_modeling_mistral.py @@ -0,0 +1,385 @@ +# coding=utf-8 +# Copyright 2023 Mistral AI and The HuggingFace Inc. team. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +""" Testing suite for the PyTorch Mistral model. """ + + +import unittest + +from transformers import AutoTokenizer, MistralConfig, is_torch_available +from transformers.testing_utils import require_torch, slow, torch_device + +from ...generation.test_utils import GenerationTesterMixin +from ...test_configuration_common import ConfigTester +from ...test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask +from ...test_pipeline_mixin import PipelineTesterMixin + + +if is_torch_available(): + import torch + + from transformers import ( + MistralForCausalLM, + MistralForSequenceClassification, + MistralModel, + ) + + +# Copied from transformers.tests.mistral.test_modelling_mistral.MistralModelTest with Llama->Mistral +class MistralModelTester: + def __init__( + self, + parent, + batch_size=13, + seq_length=7, + is_training=True, + use_input_mask=True, + use_token_type_ids=False, + use_labels=True, + vocab_size=99, + hidden_size=32, + num_hidden_layers=2, + num_attention_heads=4, + num_key_value_heads=2, + intermediate_size=37, + hidden_act="gelu", + hidden_dropout_prob=0.1, + attention_probs_dropout_prob=0.1, + max_position_embeddings=512, + type_vocab_size=16, + type_sequence_label_size=2, + initializer_range=0.02, + num_labels=3, + num_choices=4, + pad_token_id=0, + scope=None, + ): + self.parent = parent + self.batch_size = batch_size + self.seq_length = seq_length + self.is_training = is_training + self.use_input_mask = use_input_mask + self.use_token_type_ids = use_token_type_ids + self.use_labels = use_labels + self.vocab_size = vocab_size + self.hidden_size = hidden_size + self.num_hidden_layers = num_hidden_layers + self.num_attention_heads = num_attention_heads + self.num_key_value_heads = num_key_value_heads + self.intermediate_size = intermediate_size + self.hidden_act = hidden_act + self.hidden_dropout_prob = hidden_dropout_prob + self.attention_probs_dropout_prob = attention_probs_dropout_prob + self.max_position_embeddings = max_position_embeddings + self.type_vocab_size = type_vocab_size + self.type_sequence_label_size = type_sequence_label_size + self.initializer_range = initializer_range + self.num_labels = num_labels + self.num_choices = num_choices + self.pad_token_id = pad_token_id + self.scope = scope + + def prepare_config_and_inputs(self): + input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size) + + input_mask = None + if self.use_input_mask: + input_mask = random_attention_mask([self.batch_size, self.seq_length]) + + token_type_ids = None + if self.use_token_type_ids: + token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size) + + sequence_labels = None + token_labels = None + choice_labels = None + if self.use_labels: + sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size) + token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels) + choice_labels = ids_tensor([self.batch_size], self.num_choices) + + config = self.get_config() + + return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + + def get_config(self): + return MistralConfig( + vocab_size=self.vocab_size, + hidden_size=self.hidden_size, + num_hidden_layers=self.num_hidden_layers, + num_attention_heads=self.num_attention_heads, + num_key_value_heads=self.num_key_value_heads, + intermediate_size=self.intermediate_size, + hidden_act=self.hidden_act, + hidden_dropout_prob=self.hidden_dropout_prob, + attention_probs_dropout_prob=self.attention_probs_dropout_prob, + max_position_embeddings=self.max_position_embeddings, + type_vocab_size=self.type_vocab_size, + is_decoder=False, + initializer_range=self.initializer_range, + pad_token_id=self.pad_token_id, + ) + + def create_and_check_model( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + model = MistralModel(config=config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=input_mask) + result = model(input_ids) + self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size)) + + def create_and_check_model_as_decoder( + self, + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + encoder_hidden_states, + encoder_attention_mask, + ): + config.add_cross_attention = True + model = MistralModel(config) + model.to(torch_device) + model.eval() + result = model( + input_ids, + attention_mask=input_mask, + encoder_hidden_states=encoder_hidden_states, + encoder_attention_mask=encoder_attention_mask, + ) + result = model( + input_ids, + attention_mask=input_mask, + encoder_hidden_states=encoder_hidden_states, + ) + result = model(input_ids, attention_mask=input_mask) + self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size)) + + def create_and_check_for_causal_lm( + self, + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + encoder_hidden_states, + encoder_attention_mask, + ): + model = MistralForCausalLM(config=config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=input_mask, labels=token_labels) + self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size)) + + def create_and_check_decoder_model_past_large_inputs( + self, + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + encoder_hidden_states, + encoder_attention_mask, + ): + config.is_decoder = True + config.add_cross_attention = True + model = MistralForCausalLM(config=config) + model.to(torch_device) + model.eval() + + # first forward pass + outputs = model( + input_ids, + attention_mask=input_mask, + encoder_hidden_states=encoder_hidden_states, + encoder_attention_mask=encoder_attention_mask, + use_cache=True, + ) + past_key_values = outputs.past_key_values + + # create hypothetical multiple next token and extent to next_input_ids + next_tokens = ids_tensor((self.batch_size, 3), config.vocab_size) + next_mask = ids_tensor((self.batch_size, 3), vocab_size=2) + + # append to next input_ids and + next_input_ids = torch.cat([input_ids, next_tokens], dim=-1) + next_attention_mask = torch.cat([input_mask, next_mask], dim=-1) + + output_from_no_past = model( + next_input_ids, + attention_mask=next_attention_mask, + encoder_hidden_states=encoder_hidden_states, + encoder_attention_mask=encoder_attention_mask, + output_hidden_states=True, + )["hidden_states"][0] + output_from_past = model( + next_tokens, + attention_mask=next_attention_mask, + encoder_hidden_states=encoder_hidden_states, + encoder_attention_mask=encoder_attention_mask, + past_key_values=past_key_values, + output_hidden_states=True, + )["hidden_states"][0] + + # select random slice + random_slice_idx = ids_tensor((1,), output_from_past.shape[-1]).item() + output_from_no_past_slice = output_from_no_past[:, -3:, random_slice_idx].detach() + output_from_past_slice = output_from_past[:, :, random_slice_idx].detach() + + self.parent.assertTrue(output_from_past_slice.shape[1] == next_tokens.shape[1]) + + # test that outputs are equal for slice + self.parent.assertTrue(torch.allclose(output_from_past_slice, output_from_no_past_slice, atol=1e-3)) + + def prepare_config_and_inputs_for_common(self): + config_and_inputs = self.prepare_config_and_inputs() + ( + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + ) = config_and_inputs + inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask} + return config, inputs_dict + + +@require_torch +class MistralModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin, unittest.TestCase): + all_model_classes = ( + (MistralModel, MistralForCausalLM, MistralForSequenceClassification) if is_torch_available() else () + ) + all_generative_model_classes = (MistralForCausalLM,) if is_torch_available() else () + pipeline_model_mapping = ( + { + "feature-extraction": MistralModel, + "text-classification": MistralForSequenceClassification, + "text-generation": MistralForCausalLM, + "zero-shot": MistralForSequenceClassification, + } + if is_torch_available() + else {} + ) + test_headmasking = False + test_pruning = False + + def setUp(self): + self.model_tester = MistralModelTester(self) + self.config_tester = ConfigTester(self, config_class=MistralConfig, hidden_size=37) + + def test_config(self): + self.config_tester.run_common_tests() + + def test_model(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_model(*config_and_inputs) + + def test_model_various_embeddings(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + for type in ["absolute", "relative_key", "relative_key_query"]: + config_and_inputs[0].position_embedding_type = type + self.model_tester.create_and_check_model(*config_and_inputs) + + def test_Mistral_sequence_classification_model(self): + config, input_dict = self.model_tester.prepare_config_and_inputs_for_common() + print(config) + config.num_labels = 3 + input_ids = input_dict["input_ids"] + attention_mask = input_ids.ne(1).to(torch_device) + sequence_labels = ids_tensor([self.model_tester.batch_size], self.model_tester.type_sequence_label_size) + model = MistralForSequenceClassification(config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=attention_mask, labels=sequence_labels) + self.assertEqual(result.logits.shape, (self.model_tester.batch_size, self.model_tester.num_labels)) + + def test_Mistral_sequence_classification_model_for_single_label(self): + config, input_dict = self.model_tester.prepare_config_and_inputs_for_common() + config.num_labels = 3 + config.problem_type = "single_label_classification" + input_ids = input_dict["input_ids"] + attention_mask = input_ids.ne(1).to(torch_device) + sequence_labels = ids_tensor([self.model_tester.batch_size], self.model_tester.type_sequence_label_size) + model = MistralForSequenceClassification(config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=attention_mask, labels=sequence_labels) + self.assertEqual(result.logits.shape, (self.model_tester.batch_size, self.model_tester.num_labels)) + + def test_Mistral_sequence_classification_model_for_multi_label(self): + config, input_dict = self.model_tester.prepare_config_and_inputs_for_common() + config.num_labels = 3 + config.problem_type = "multi_label_classification" + input_ids = input_dict["input_ids"] + attention_mask = input_ids.ne(1).to(torch_device) + sequence_labels = ids_tensor( + [self.model_tester.batch_size, config.num_labels], self.model_tester.type_sequence_label_size + ).to(torch.float) + model = MistralForSequenceClassification(config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=attention_mask, labels=sequence_labels) + self.assertEqual(result.logits.shape, (self.model_tester.batch_size, self.model_tester.num_labels)) + + @unittest.skip("Mistral buffers include complex numbers, which breaks this test") + def test_save_load_fast_init_from_base(self): + pass + + @unittest.skip("Mistral uses GQA on all models so the KV cache is a non standard format") + def test_past_key_values_format(self): + pass + + +@require_torch +class MistralIntegrationTest(unittest.TestCase): + @slow + def test_model_7b_logits(self): + input_ids = [1, 306, 4658, 278, 6593, 310, 2834, 338] + model = MistralForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1", device_map="auto") + out = model(torch.tensor([input_ids])).logits + # Expected mean on dim = -1 + EXPECTED_MEAN = torch.tensor([[-2.5548, -2.5737, -3.0600, -2.5906, -2.8478, -2.8118, -2.9325, -2.7694]]) + torch.testing.assert_close(out.mean(-1), EXPECTED_MEAN, atol=1e-2, rtol=1e-2) + # slicing logits[0, 0, 0:30] + # fmt: off + EXPECTED_SLICE = torch.tensor([-5.8781, -5.8616, -0.1052, -4.7200, -5.8781, -5.8774, -5.8773, -5.8777, -5.8781, -5.8780, -5.8781, -5.8779, -1.0787, 1.7583, -5.8779, -5.8780, -5.8783, -5.8778, -5.8776, -5.8781, -5.8784, -5.8778, -5.8778, -5.8777, -5.8779, -5.8778, -5.8776, -5.8780, -5.8779, -5.8781]) + # fmt: on + print(out[0, 0, :30]) + torch.testing.assert_close(out[0, 0, :30], EXPECTED_SLICE, atol=1e-4, rtol=1e-4) + + @slow + def test_model_7b_generation(self): + EXPECTED_TEXT_COMPLETION = ( + """My favourite condiment is mayonnaise. I love it on sandwiches, in salads, on burgers""" + ) + prompt = "My favourite condiment is " + tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-v0.1", use_fast=False) + input_ids = tokenizer.encode(prompt, return_tensors="pt").to(torch_device) + model = MistralForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1").to(torch_device) + + # greedy generation outputs + generated_ids = model.generate(input_ids, max_new_tokens=20, temperature=0) + text = tokenizer.decode(generated_ids[0], skip_special_tokens=True) + self.assertEqual(EXPECTED_TEXT_COMPLETION, text) diff --git a/utils/not_doctested.txt b/utils/not_doctested.txt index 969af2ebfb..edd9469439 100644 --- a/utils/not_doctested.txt +++ b/utils/not_doctested.txt @@ -665,6 +665,7 @@ src/transformers/models/megatron_gpt2/checkpoint_reshaping_and_interoperability. src/transformers/models/megatron_gpt2/convert_megatron_gpt2_checkpoint.py src/transformers/models/mgp_str/configuration_mgp_str.py src/transformers/models/mgp_str/modeling_mgp_str.py +src/transformers/models/mistral/configuration_mistral.py src/transformers/models/mluke/convert_mluke_original_pytorch_checkpoint_to_pytorch.py src/transformers/models/mobilebert/convert_mobilebert_original_tf_checkpoint_to_pytorch.py src/transformers/models/mobilenet_v1/configuration_mobilenet_v1.py