phi0112358 commited on
Commit
b4496ac
1 Parent(s): d22e6cf

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +173 -0
README.md CHANGED
@@ -1,3 +1,176 @@
1
  ---
 
 
 
 
 
 
2
  license: apache-2.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ pipeline_tag: sentence-similarity
3
+ tags:
4
+ - sentence-transformers
5
+ - feature-extraction
6
+ - sentence-similarity
7
+ language: en
8
  license: apache-2.0
9
+ datasets:
10
+ - s2orc
11
+ - flax-sentence-embeddings/stackexchange_xml
12
+ - ms_marco
13
+ - gooaq
14
+ - yahoo_answers_topics
15
+ - code_search_net
16
+ - search_qa
17
+ - eli5
18
+ - snli
19
+ - multi_nli
20
+ - wikihow
21
+ - natural_questions
22
+ - trivia_qa
23
+ - embedding-data/sentence-compression
24
+ - embedding-data/flickr30k-captions
25
+ - embedding-data/altlex
26
+ - embedding-data/simple-wiki
27
+ - embedding-data/QQP
28
+ - embedding-data/SPECTER
29
+ - embedding-data/PAQ_pairs
30
+ - embedding-data/WikiAnswers
31
+ library_name: sentence-transformers
32
  ---
33
+
34
+
35
+ # all-MiniLM-L6-v2-GGML
36
+ This 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.
37
+
38
+ ## Usage (Sentence-Transformers)
39
+ Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:
40
+
41
+ ```
42
+ pip install -U sentence-transformers
43
+ ```
44
+
45
+ Then you can use the model like this:
46
+ ```python
47
+ from sentence_transformers import SentenceTransformer
48
+ sentences = ["This is an example sentence", "Each sentence is converted"]
49
+
50
+ model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
51
+ embeddings = model.encode(sentences)
52
+ print(embeddings)
53
+ ```
54
+
55
+ ## Usage (HuggingFace Transformers)
56
+ Without [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.
57
+
58
+ ```python
59
+ from transformers import AutoTokenizer, AutoModel
60
+ import torch
61
+ import torch.nn.functional as F
62
+
63
+ #Mean Pooling - Take attention mask into account for correct averaging
64
+ def mean_pooling(model_output, attention_mask):
65
+ token_embeddings = model_output[0] #First element of model_output contains all token embeddings
66
+ input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
67
+ return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
68
+
69
+
70
+ # Sentences we want sentence embeddings for
71
+ sentences = ['This is an example sentence', 'Each sentence is converted']
72
+
73
+ # Load model from HuggingFace Hub
74
+ tokenizer = AutoTokenizer.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')
75
+ model = AutoModel.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')
76
+
77
+ # Tokenize sentences
78
+ encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
79
+
80
+ # Compute token embeddings
81
+ with torch.no_grad():
82
+ model_output = model(**encoded_input)
83
+
84
+ # Perform pooling
85
+ sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
86
+
87
+ # Normalize embeddings
88
+ sentence_embeddings = F.normalize(sentence_embeddings, p=2, dim=1)
89
+
90
+ print("Sentence embeddings:")
91
+ print(sentence_embeddings)
92
+ ```
93
+
94
+ ## Evaluation Results
95
+
96
+ For an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: [https://seb.sbert.net](https://seb.sbert.net?model_name=sentence-transformers/all-MiniLM-L6-v2)
97
+
98
+ ------
99
+
100
+ ## Background
101
+
102
+ The project aims to train sentence embedding models on very large sentence level datasets using a self-supervised
103
+ contrastive 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
104
+ 1B 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.
105
+
106
+ We developped this model during the
107
+ [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),
108
+ organized by Hugging Face. We developped this model as part of the project:
109
+ [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.
110
+
111
+ ## Intended uses
112
+
113
+ Our model is intented to be used as a sentence and short paragraph encoder. Given an input text, it ouptuts a vector which captures
114
+ the semantic information. The sentence vector may be used for information retrieval, clustering or sentence similarity tasks.
115
+
116
+ By default, input text longer than 256 word pieces is truncated.
117
+
118
+
119
+ ## Training procedure
120
+
121
+ ### Pre-training
122
+
123
+ We 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.
124
+
125
+ ### Fine-tuning
126
+
127
+ We fine-tune the model using a contrastive objective. Formally, we compute the cosine similarity from each possible sentence pairs from the batch.
128
+ We then apply the cross entropy loss by comparing with true pairs.
129
+
130
+ #### Hyper parameters
131
+
132
+ We trained ou model on a TPU v3-8. We train the model during 100k steps using a batch size of 1024 (128 per TPU core).
133
+ We use a learning rate warm up of 500. The sequence length was limited to 128 tokens. We used the AdamW optimizer with
134
+ a 2e-5 learning rate. The full training script is accessible in this current repository: `train_script.py`.
135
+
136
+ #### Training data
137
+
138
+ We use the concatenation from multiple datasets to fine-tune our model. The total number of sentence pairs is above 1 billion sentences.
139
+ We sampled each dataset given a weighted probability which configuration is detailed in the `data_config.json` file.
140
+
141
+
142
+ | Dataset | Paper | Number of training tuples |
143
+ |--------------------------------------------------------|:----------------------------------------:|:--------------------------:|
144
+ | [Reddit comments (2015-2018)](https://github.com/PolyAI-LDN/conversational-datasets/tree/master/reddit) | [paper](https://arxiv.org/abs/1904.06472) | 726,484,430 |
145
+ | [S2ORC](https://github.com/allenai/s2orc) Citation pairs (Abstracts) | [paper](https://aclanthology.org/2020.acl-main.447/) | 116,288,806 |
146
+ | [WikiAnswers](https://github.com/afader/oqa#wikianswers-corpus) Duplicate question pairs | [paper](https://doi.org/10.1145/2623330.2623677) | 77,427,422 |
147
+ | [PAQ](https://github.com/facebookresearch/PAQ) (Question, Answer) pairs | [paper](https://arxiv.org/abs/2102.07033) | 64,371,441 |
148
+ | [S2ORC](https://github.com/allenai/s2orc) Citation pairs (Titles) | [paper](https://aclanthology.org/2020.acl-main.447/) | 52,603,982 |
149
+ | [S2ORC](https://github.com/allenai/s2orc) (Title, Abstract) | [paper](https://aclanthology.org/2020.acl-main.447/) | 41,769,185 |
150
+ | [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) (Title, Body) pairs | - | 25,316,456 |
151
+ | [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) (Title+Body, Answer) pairs | - | 21,396,559 |
152
+ | [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) (Title, Answer) pairs | - | 21,396,559 |
153
+ | [MS MARCO](https://microsoft.github.io/msmarco/) triplets | [paper](https://doi.org/10.1145/3404835.3462804) | 9,144,553 |
154
+ | [GOOAQ: Open Question Answering with Diverse Answer Types](https://github.com/allenai/gooaq) | [paper](https://arxiv.org/pdf/2104.08727.pdf) | 3,012,496 |
155
+ | [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 |
156
+ | [Code Search](https://huggingface.co/datasets/code_search_net) | - | 1,151,414 |
157
+ | [COCO](https://cocodataset.org/#home) Image captions | [paper](https://link.springer.com/chapter/10.1007%2F978-3-319-10602-1_48) | 828,395|
158
+ | [SPECTER](https://github.com/allenai/specter) citation triplets | [paper](https://doi.org/10.18653/v1/2020.acl-main.207) | 684,100 |
159
+ | [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 |
160
+ | [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 |
161
+ | [SearchQA](https://huggingface.co/datasets/search_qa) | [paper](https://arxiv.org/abs/1704.05179) | 582,261 |
162
+ | [Eli5](https://huggingface.co/datasets/eli5) | [paper](https://doi.org/10.18653/v1/p19-1346) | 325,475 |
163
+ | [Flickr 30k](https://shannon.cs.illinois.edu/DenotationGraph/) | [paper](https://transacl.org/ojs/index.php/tacl/article/view/229/33) | 317,695 |
164
+ | [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) Duplicate questions (titles) | | 304,525 |
165
+ | 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 |
166
+ | [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) Duplicate questions (bodies) | | 250,519 |
167
+ | [Stack Exchange](https://huggingface.co/datasets/flax-sentence-embeddings/stackexchange_xml) Duplicate questions (titles+bodies) | | 250,460 |
168
+ | [Sentence Compression](https://github.com/google-research-datasets/sentence-compression) | [paper](https://www.aclweb.org/anthology/D13-1155/) | 180,000 |
169
+ | [Wikihow](https://github.com/pvl/wikihow_pairs_dataset) | [paper](https://arxiv.org/abs/1810.09305) | 128,542 |
170
+ | [Altlex](https://github.com/chridey/altlex/) | [paper](https://aclanthology.org/P16-1135.pdf) | 112,696 |
171
+ | [Quora Question Triplets](https://quoradata.quora.com/First-Quora-Dataset-Release-Question-Pairs) | - | 103,663 |
172
+ | [Simple Wikipedia](https://cs.pomona.edu/~dkauchak/simplification/) | [paper](https://www.aclweb.org/anthology/P11-2117/) | 102,225 |
173
+ | [Natural Questions (NQ)](https://ai.google.com/research/NaturalQuestions) | [paper](https://transacl.org/ojs/index.php/tacl/article/view/1455) | 100,231 |
174
+ | [SQuAD2.0](https://rajpurkar.github.io/SQuAD-explorer/) | [paper](https://aclanthology.org/P18-2124.pdf) | 87,599 |
175
+ | [TriviaQA](https://huggingface.co/datasets/trivia_qa) | - | 73,346 |
176
+ | **Total** | | **1,170,060,424** |