YUANHENG666/MNLP_M3_document_encoder

1---2language: en3license: apache-2.04library_name: sentence-transformers5tags:6- sentence-transformers7- feature-extraction8- sentence-similarity9- transformers10datasets:11- s2orc12- flax-sentence-embeddings/stackexchange_xml13- ms_marco14- gooaq15- yahoo_answers_topics16- code_search_net17- search_qa18- eli519- snli20- multi_nli21- wikihow22- natural_questions23- trivia_qa24- embedding-data/sentence-compression25- embedding-data/flickr30k-captions26- embedding-data/altlex27- embedding-data/simple-wiki28- embedding-data/QQP29- embedding-data/SPECTER30- embedding-data/PAQ_pairs31- embedding-data/WikiAnswers32pipeline_tag: sentence-similarity33---34 35 36# all-MiniLM-L6-v237This is a [sentence-transformers](https://www.SBERT.net) model: It maps sentences & paragraphs to a 384 dimensional dense vector space and can be used for tasks like clustering or semantic search.38 39## Usage (Sentence-Transformers)40Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:41 42```43pip install -U sentence-transformers44```45 46Then you can use the model like this:47```python48from sentence_transformers import SentenceTransformer49sentences = ["This is an example sentence", "Each sentence is converted"]50 51model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')52embeddings = model.encode(sentences)53print(embeddings)54```55 56## Usage (HuggingFace Transformers)57Without [sentence-transformers](https://www.SBERT.net), you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.58 59```python60from transformers import AutoTokenizer, AutoModel61import torch62import torch.nn.functional as F63 64#Mean Pooling - Take attention mask into account for correct averaging65def mean_pooling(model_output, attention_mask):66    token_embeddings = model_output[0] #First element of model_output contains all token embeddings67    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()68    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)69 70 71# Sentences we want sentence embeddings for72sentences = ['This is an example sentence', 'Each sentence is converted']73 74# Load model from HuggingFace Hub75tokenizer = AutoTokenizer.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')76model = AutoModel.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')77 78# Tokenize sentences79encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')80 81# Compute token embeddings82with torch.no_grad():83    model_output = model(**encoded_input)84 85# Perform pooling86sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])87 88# Normalize embeddings89sentence_embeddings = F.normalize(sentence_embeddings, p=2, dim=1)90 91print("Sentence embeddings:")92print(sentence_embeddings)93```94 95------96 97## Background98 99The project aims to train sentence embedding models on very large sentence level datasets using a self-supervised 100contrastive learning objective. We used the pretrained [`nreimers/MiniLM-L6-H384-uncased`](https://huggingface.co/nreimers/MiniLM-L6-H384-uncased) model and fine-tuned in on a 1011B sentence pairs dataset. We use a contrastive learning objective: given a sentence from the pair, the model should predict which out of a set of randomly sampled other sentences, was actually paired with it in our dataset.102 103We developed this model during the 104[Community week using JAX/Flax for NLP & CV](https://discuss.huggingface.co/t/open-to-the-community-community-week-using-jax-flax-for-nlp-cv/7104), 105organized by Hugging Face. We developed this model as part of the project:106[Train the Best Sentence Embedding Model Ever with 1B Training Pairs](https://discuss.huggingface.co/t/train-the-best-sentence-embedding-model-ever-with-1b-training-pairs/7354). We benefited from efficient hardware infrastructure to run the project: 7 TPUs v3-8, as well as intervention from Googles Flax, JAX, and Cloud team member about efficient deep learning frameworks.107 108## Intended uses109 110Our model is intended to be used as a sentence and short paragraph encoder. Given an input text, it outputs a vector which captures 111the semantic information. The sentence vector may be used for information retrieval, clustering or sentence similarity tasks.112 113By default, input text longer than 256 word pieces is truncated.114 115 116## Training procedure117 118### Pre-training 119 120We use the pretrained [`nreimers/MiniLM-L6-H384-uncased`](https://huggingface.co/nreimers/MiniLM-L6-H384-uncased) model. Please refer to the model card for more detailed information about the pre-training procedure.121 122### Fine-tuning 123 124We fine-tune the model using a contrastive objective. Formally, we compute the cosine similarity from each possible sentence pairs from the batch.125We then apply the cross entropy loss by comparing with true pairs.126 127#### Hyper parameters128 129We trained our model on a TPU v3-8. We train the model during 100k steps using a batch size of 1024 (128 per TPU core).130We use a learning rate warm up of 500. The sequence length was limited to 128 tokens. We used the AdamW optimizer with131a 2e-5 learning rate. The full training script is accessible in this current repository: `train_script.py`.132 133#### Training data134 135We use the concatenation from multiple datasets to fine-tune our model. The total number of sentence pairs is above 1 billion sentences.136We sampled each dataset given a weighted probability which configuration is detailed in the `data_config.json` file.137 138 139| Dataset                                                  | Paper                                    | Number of training tuples  |140|--------------------------------------------------------|:----------------------------------------:|:--------------------------:|141| [Reddit comments (2015-2018)](https://github.com/PolyAI-LDN/conversational-datasets/tree/master/reddit) | [paper](https://arxiv.org/abs/1904.06472) | 726,484,430 |142| [S2ORC](https://github.com/allenai/s2orc) Citation pairs (Abstracts) | [paper](https://aclanthology.org/2020.acl-main.447/) | 116,288,806 |143| [WikiAnswers](https://github.com/afader/oqa#wikianswers-corpus) Duplicate question pairs | [paper](https://doi.org/10.1145/2623330.2623677) | 77,427,422 |144| [PAQ](https://github.com/facebookresearch/PAQ) (Question, Answer) pairs | [paper](https://arxiv.org/abs/2102.07033) | 64,371,441 |145| [S2ORC](https://github.com/allenai/s2orc) Citation pairs (Titles) | [paper](https://aclanthology.org/2020.acl-main.447/) | 52,603,982 |146| [S2ORC](https://github.com/allenai/s2orc) (Title, Abstract) | [paper](https://aclanthology.org/2020.acl-main.447/) | 41,769,185 |147| [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) (Title, Body) pairs  | - | 25,316,456 |148| [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) (Title+Body, Answer) pairs  | - | 21,396,559 |149| [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) (Title, Answer) pairs  | - | 21,396,559 |150| [MS MARCO](https://microsoft.github.io/msmarco/) triplets | [paper](https://doi.org/10.1145/3404835.3462804) | 9,144,553 |151| [GOOAQ: Open Question Answering with Diverse Answer Types](https://github.com/allenai/gooaq) | [paper](https://arxiv.org/pdf/2104.08727.pdf) | 3,012,496 |152| [Yahoo Answers](https://www.kaggle.com/soumikrakshit/yahoo-answers-dataset) (Title, Answer) | [paper](https://proceedings.neurips.cc/paper/2015/hash/250cf8b51c773f3f8dc8b4be867a9a02-Abstract.html) | 1,198,260 |153| [Code Search](https://huggingface.co/datasets/code_search_net) | - | 1,151,414 |154| [COCO](https://cocodataset.org/#home) Image captions | [paper](https://link.springer.com/chapter/10.1007%2F978-3-319-10602-1_48) | 828,395|155| [SPECTER](https://github.com/allenai/specter) citation triplets | [paper](https://doi.org/10.18653/v1/2020.acl-main.207) | 684,100 |156| [Yahoo Answers](https://www.kaggle.com/soumikrakshit/yahoo-answers-dataset) (Question, Answer) | [paper](https://proceedings.neurips.cc/paper/2015/hash/250cf8b51c773f3f8dc8b4be867a9a02-Abstract.html) | 681,164 |157| [Yahoo Answers](https://www.kaggle.com/soumikrakshit/yahoo-answers-dataset) (Title, Question) | [paper](https://proceedings.neurips.cc/paper/2015/hash/250cf8b51c773f3f8dc8b4be867a9a02-Abstract.html) | 659,896 |158| [SearchQA](https://huggingface.co/datasets/search_qa) | [paper](https://arxiv.org/abs/1704.05179) | 582,261 |159| [Eli5](https://huggingface.co/datasets/eli5) | [paper](https://doi.org/10.18653/v1/p19-1346) | 325,475 |160| [Flickr 30k](https://shannon.cs.illinois.edu/DenotationGraph/) | [paper](https://transacl.org/ojs/index.php/tacl/article/view/229/33) | 317,695 |161| [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) Duplicate questions (titles) | | 304,525 |162| AllNLI ([SNLI](https://nlp.stanford.edu/projects/snli/) and [MultiNLI](https://cims.nyu.edu/~sbowman/multinli/) | [paper SNLI](https://doi.org/10.18653/v1/d15-1075), [paper MultiNLI](https://doi.org/10.18653/v1/n18-1101) | 277,230 | 163| [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) Duplicate questions (bodies) | | 250,519 |164| [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) Duplicate questions (titles+bodies) | | 250,460 |165| [Sentence Compression](https://github.com/google-research-datasets/sentence-compression) | [paper](https://www.aclweb.org/anthology/D13-1155/) | 180,000 |166| [Wikihow](https://github.com/pvl/wikihow_pairs_dataset) | [paper](https://arxiv.org/abs/1810.09305) | 128,542 |167| [Altlex](https://github.com/chridey/altlex/) | [paper](https://aclanthology.org/P16-1135.pdf) | 112,696 |168| [Quora Question Triplets](https://quoradata.quora.com/First-Quora-Dataset-Release-Question-Pairs) | - | 103,663 |169| [Simple Wikipedia](https://cs.pomona.edu/~dkauchak/simplification/) | [paper](https://www.aclweb.org/anthology/P11-2117/) | 102,225 |170| [Natural Questions (NQ)](https://ai.google.com/research/NaturalQuestions) | [paper](https://transacl.org/ojs/index.php/tacl/article/view/1455) | 100,231 |171| [SQuAD2.0](https://rajpurkar.github.io/SQuAD-explorer/) | [paper](https://aclanthology.org/P18-2124.pdf) | 87,599 |172| [TriviaQA](https://huggingface.co/datasets/trivia_qa) | - | 73,346 |173| **Total** | | **1,170,060,424** |