sentence_similarity_spanish_es

Introduction

The sentence_similarity_spanish_es model, developed by HIIAMSID, is a Spanish sentence-transformers model that maps sentences and paragraphs to a 768-dimensional dense vector space. It is designed for tasks such as clustering or semantic search.

Architecture

The model is built using a Transformer architecture with BERT as the underlying model. It includes a pooling layer that supports mean pooling of token embeddings. The model is configured with a maximum sequence length of 512 and does not apply lower casing.

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, '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})
)

Training

The model was trained using the stsb_multi_mt dataset with dccuchile/bert-base-spanish-wwm-cased as the base model. Training involved a CosineSimilarityLoss with a batch size of 16, over 4 epochs. An AdamW optimizer with a learning rate of 2e-05 and a warmup-linear scheduler was used.

{
    "epochs": 4,
    "evaluation_steps": 1000,
    "optimizer_params": {
        "lr": 2e-05
    },
    "warmup_steps": 144,
    "weight_decay": 0.01
}

Guide: Running Locally

Using Sentence-Transformers

1. Install the sentence-transformers library:

   pip install -U sentence-transformers
   

2. Use the model for generating embeddings:

   from sentence_transformers import SentenceTransformer
   
   sentences = ['Mi nombre es Siddhartha', 'Mis amigos me llamaron por mi nombre Siddhartha']
   model = SentenceTransformer('hiiamsid/sentence_similarity_spanish_es')
   embeddings = model.encode(sentences)
   print(embeddings)
   

Using Hugging Face Transformers

1. Install the necessary libraries:

   pip install transformers torch
   

2. Use the model with pooling for sentence embeddings:

   from transformers import AutoTokenizer, AutoModel
   import torch
   
   def mean_pooling(model_output, attention_mask):
       token_embeddings = model_output[0]
       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 = ['Mi nombre es Siddhartha', 'Mis amigos me llamaron por mi nombre Siddhartha']
   tokenizer = AutoTokenizer.from_pretrained('hiiamsid/sentence_similarity_spanish_es')
   model = AutoModel.from_pretrained('hiiamsid/sentence_similarity_spanish_es')
   
   encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
   with torch.no_grad():
       model_output = model(**encoded_input)
   
   sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
   print("Sentence embeddings:")
   print(sentence_embeddings)
   

Cloud GPUs

For efficient processing, it is recommended to use cloud-based GPUs from providers like AWS, Google Cloud, or Azure to speed up inference and training processes.

License

This model is licensed under the Apache-2.0 License, which allows for both personal and commercial use.