stjiris/bert-large-portuguese-cased-legal-mlm-nli-sts-v1

Usar este modelo es fácil cuando tienes sentence-transformers instalado:

pip install -U sentence-transformers

Luego puedes usar el modelo así:

from sentence_transformers import SentenceTransformer
sentences = ["Isto é um exemplo", "Isto é um outro exemplo"]

model = SentenceTransformer('stjiris/bert-large-portuguese-cased-legal-mlm-nli-sts-v1')
embeddings = model.encode(sentences)
print(embeddings)

Usar (HuggingFace Transformers)

from transformers import AutoTokenizer, AutoModel
import torch

# Mean Pooling - Tomar en cuenta la máscara de atención para un promedio correcto
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] # El primer elemento de model_output contiene todos los 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)

# Oraciones para las cuales queremos los embeddings de oraciones
sentences = ['This is an example sentence', 'Each sentence is converted']

# Cargar el modelo desde HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('stjiris/bert-large-portuguese-cased-legal-mlm-nli-sts-v1')
model = AutoModel.from_pretrained('stjiris/bert-large-portuguese-cased-legal-mlm-nli-sts-v1')

# Tokenizar oraciones
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

# Calcular embeddings de tokens
with torch.no_grad():
    model_output = model(**encoded_input)
# Realizar pooling. En este caso, pooling de media.
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
print("Embeddings de oraciones:")
print(sentence_embeddings)

Arquitectura completa del modelo

SentenceTransformer(
(0): Transformer({'max_seq_length': 514, 'do_lower_case': False}) with Transformer model: BertModel 
(1): Pooling({'word_embedding_dimension': 1028, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False})
)