pipeline_tag: sentence-similarity
tags:
- sentence-transformers
- feature-extraction
- sentence-similarity
- transformers
widget:
- source_sentence: This is a Norwegian boy
sentences:
- Dette er en norsk gutt
- This is an English boy
- This is a dog
example_title: Cross Language
- source_sentence: Det er noen dyr utenfor vinduet
sentences:
- På utsiden kan jeg høre noen hunder
- Noen mennesker prater utenfor vinduet
- Alle burde ha kjæledyr
example_title: Paraphrases
- source_sentence: En kvinne sitter i en stol
sentences:
- A woman is sitting in a chair
- Hun slapper av og leser i en bok
- Hun løper maraton
example_title: Paraphrases across language
nb-sbert
This is a sentence-transformers model trained on the machine translated mnli-dataset starting from nb-bert-base.
The model maps sentences & paragraphs to a 768 dimensional dense vector space. This vector can be used for tasks like clustering and semantic search. Below we give some examples on how to use the model in different framework. The easiest way is to simply measure the cosine distance between two sentences. Sentences that are close to each other in meaning, will have a small cosine distance and a similarity close to 1. The model is trained in a way where we try to keep this distance also between languages. Ideally a English-Norwegian sentence pair should have high similarity.
Usage (Sentence-Transformers)
Using this model becomes easy when you have sentence-transformers installed:
pip install -U sentence-transformers
Then you can use the model like this:
from sentence_transformers import SentenceTransformer, util
sentences = ["This is a Norwegian boy", "Dette er en norsk gutt"]
model = SentenceTransformer('NbAiLab/nb-sbert')
embeddings = model.encode(sentences)
print(embeddings)
# Compute cosine-similarities with sentence transformers
cosine_scores = util.cos_sim(embeddings[0],embeddings[1])
print(cosine_scores)
# Compute cosine-similarities with SciPy
from scipy import spatial
scipy_cosine_scores = 1 - spatial.distance.cosine(embeddings[0],embeddings[1])
print(scipy_cosine_scores)
# Both should give 0.8250 in the example above.
Usage (HuggingFace Transformers)
Without sentence-transformers, 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.
from transformers import AutoTokenizer, AutoModel
import torch
#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
token_embeddings = model_output[0] #First element of model_output contains all token embeddings
input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
# Sentences we want sentence embeddings for
sentences = ["This is a Norwegian boy", "Dette er en norsk gutt"]
# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('NbAiLab/nb-sbert')
model = AutoModel.from_pretrained('NbAiLab/nb-sbert')
# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
# Compute token embeddings
with torch.no_grad():
model_output = model(**encoded_input)
# Perform pooling. In this case, mean pooling.
embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
print(embeddings)
# Compute cosine-similarities with SciPy
from scipy import spatial
scipy_cosine_scores = 1 - spatial.distance.cosine(embeddings[0],embeddings[1])
print(scipy_cosine_scores)
# This should give 0.8250 in the example above.
Training
The model was trained with the parameters:
DataLoader:
sentence_transformers.datasets.NoDuplicatesDataLoader.NoDuplicatesDataLoader of length 16471 with parameters:
{'batch_size': 32}
Loss:
sentence_transformers.losses.MultipleNegativesRankingLoss.MultipleNegativesRankingLoss with parameters:
{'scale': 20.0, 'similarity_fct': 'cos_sim'}
Parameters of the fit()-Method:
{
"epochs": 1,
"evaluation_steps": 1647,
"evaluator": "sentence_transformers.evaluation.EmbeddingSimilarityEvaluator.EmbeddingSimilarityEvaluator",
"max_grad_norm": 1,
"optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
"optimizer_params": {
"lr": 2e-05
},
"scheduler": "WarmupLinear",
"steps_per_epoch": null,
"warmup_steps": 1648,
"weight_decay": 0.01
}
Full Model Architecture
SentenceTransformer(
(0): Transformer({'max_seq_length': 75, 'do_lower_case': False}) with Transformer model: BertModel
(1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
)