mlsa-iai-msu-lab/sci-rus-tiny

from transformers import AutoTokenizer, AutoModel
import torch.nn.functional as F
import torch

tokenizer = AutoTokenizer.from_pretrained("mlsa-iai-msu-lab/sci-rus-tiny")
model = AutoModel.from_pretrained("mlsa-iai-msu-lab/sci-rus-tiny")

# model.cuda()  # si quieres usar una GPU

def mean_pooling(model_output, attention_mask):
  token_embeddings = model_output[0] # El primer elemento de model_output contiene todos los embeddings de tokens
  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)

def get_sentence_embedding(title, abstract, model, tokenizer, max_length=None):
  # Tokenizar frases
  sentence = ''.join([title, abstract])
  encoded_input = tokenizer([sentence], padding=True, truncation=True, return_tensors='pt', max_length=max_length).to(model.device)
  # Calcular embeddings de tokens
  with torch.no_grad():
    model_output = model(**encoded_input)
  # Realizar pooling
  sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
  # Normalizar embeddings
  sentence_embeddings = F.normalize(sentence_embeddings, p=2, dim=1)
  return sentence_embeddings.cpu().detach().numpy()[0]

print(get_sentence_embedding('some title', 'some abstract', model, tokenizer).shape)
# (312,)

# También puedes usar sentence_transformers:
from sentence_transformers import SentenceTransformer

model = SentenceTransformer('mlsa-iai-msu-lab/sci-rus-tiny')
embeddings = model.encode(['some title' + '' + 'some abstract'])
print(embeddings[0].shape)
# (312,)