bge-micro
This is a sentence-transformers model: It maps sentences & paragraphs to a 384 dimensional dense vector space and can be used for tasks like clustering or semantic search. It is distilled from bge-small-en-v1.5, with 1/4 the non-embedding parameters. It has 1/2 the parameters of the smallest commonly-used embedding model, all-MiniLM-L6-v2, with similar performance.
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
sentences = ["This is an example sentence", "Each sentence is converted"]
model = SentenceTransformer('{MODEL_NAME}')
embeddings = model.encode(sentences)
print(embeddings)
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 an example sentence', 'Each sentence is converted']
# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('{MODEL_NAME}')
model = AutoModel.from_pretrained('{MODEL_NAME}')
# 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.
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
print("Sentence embeddings:")
print(sentence_embeddings)
Evaluation Results
For an automated evaluation of this model, see the Sentence Embeddings Benchmark: https://seb.sbert.net
Full Model Architecture
SentenceTransformer(
(0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel
(1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
)
Citing & Authors
- Downloads last month
- 6,106
Space using TaylorAI/bge-micro 1
Evaluation results
- accuracy on MTEB AmazonCounterfactualClassification (en)test set self-reported66.269
- ap on MTEB AmazonCounterfactualClassification (en)test set self-reported28.174
- f1 on MTEB AmazonCounterfactualClassification (en)test set self-reported59.725
- accuracy on MTEB AmazonPolarityClassificationtest set self-reported75.369
- ap on MTEB AmazonPolarityClassificationtest set self-reported69.642
- f1 on MTEB AmazonPolarityClassificationtest set self-reported75.291
- accuracy on MTEB AmazonReviewsClassification (en)test set self-reported35.806
- f1 on MTEB AmazonReviewsClassification (en)test set self-reported35.507
- map_at_1 on MTEB ArguAnatest set self-reported27.240
- map_at_10 on MTEB ArguAnatest set self-reported42.832
- map_at_100 on MTEB ArguAnatest set self-reported43.797
- map_at_1000 on MTEB ArguAnatest set self-reported43.804
- map_at_3 on MTEB ArguAnatest set self-reported38.134
- map_at_5 on MTEB ArguAnatest set self-reported40.744
- mrr_at_1 on MTEB ArguAnatest set self-reported27.952
- mrr_at_10 on MTEB ArguAnatest set self-reported43.111
- mrr_at_100 on MTEB ArguAnatest set self-reported44.083
- mrr_at_1000 on MTEB ArguAnatest set self-reported44.090
- mrr_at_3 on MTEB ArguAnatest set self-reported38.431
- mrr_at_5 on MTEB ArguAnatest set self-reported41.019
- ndcg_at_1 on MTEB ArguAnatest set self-reported27.240
- ndcg_at_10 on MTEB ArguAnatest set self-reported51.513
- ndcg_at_100 on MTEB ArguAnatest set self-reported55.762
- ndcg_at_1000 on MTEB ArguAnatest set self-reported55.938
- ndcg_at_3 on MTEB ArguAnatest set self-reported41.743
- ndcg_at_5 on MTEB ArguAnatest set self-reported46.454
- precision_at_1 on MTEB ArguAnatest set self-reported27.240
- precision_at_10 on MTEB ArguAnatest set self-reported7.930
- precision_at_100 on MTEB ArguAnatest set self-reported0.982
- precision_at_1000 on MTEB ArguAnatest set self-reported0.100
- precision_at_3 on MTEB ArguAnatest set self-reported17.402
- precision_at_5 on MTEB ArguAnatest set self-reported12.731
- recall_at_1 on MTEB ArguAnatest set self-reported27.240
- recall_at_10 on MTEB ArguAnatest set self-reported79.303
- recall_at_100 on MTEB ArguAnatest set self-reported98.151
- recall_at_1000 on MTEB ArguAnatest set self-reported99.502
- recall_at_3 on MTEB ArguAnatest set self-reported52.205
- recall_at_5 on MTEB ArguAnatest set self-reported63.656
- v_measure on MTEB ArxivClusteringP2Ptest set self-reported44.598
- v_measure on MTEB ArxivClusteringS2Stest set self-reported34.480
- map on MTEB AskUbuntuDupQuestionstest set self-reported58.093
- mrr on MTEB AskUbuntuDupQuestionstest set self-reported72.184
- cos_sim_pearson on MTEB BIOSSEStest set self-reported85.476
- cos_sim_spearman on MTEB BIOSSEStest set self-reported83.416
- euclidean_pearson on MTEB BIOSSEStest set self-reported84.220
- euclidean_spearman on MTEB BIOSSEStest set self-reported83.466
- manhattan_pearson on MTEB BIOSSEStest set self-reported83.831
- manhattan_spearman on MTEB BIOSSEStest set self-reported83.133
- accuracy on MTEB Banking77Classificationtest set self-reported80.581
- f1 on MTEB Banking77Classificationtest set self-reported80.533
- v_measure on MTEB BiorxivClusteringP2Ptest set self-reported37.135