# Requiere: pip install pytorch-crf
from transformers import AutoModelForTokenClassification, AutoConfig
import json

config = AutoConfig.from_pretrained("cvalenciaunivalle/bilstm-crf-fasttext-charcnn-conll-bs8", trust_remote_code=True)
model  = AutoModelForTokenClassification.from_pretrained("cvalenciaunivalle/bilstm-crf-fasttext-charcnn-conll-bs8", trust_remote_code=True)

# Cargar vocab (incluye word2idx, char2idx, id2tag)
from huggingface_hub import hf_hub_download
vocab_path = hf_hub_download("cvalenciaunivalle/bilstm-crf-fasttext-charcnn-conll-bs8", "vocab.json")
with open(vocab_path) as f:
    vocab = json.load(f)

# Predecir
oraciones_tokenizadas = [["Juan", "vive", "en", "Bogotá", "."]]
tags = model.predict(oraciones_tokenizadas, vocab)
print(tags)  # [['B-PER', 'O', 'O', 'B-LOC', 'O']]

Detalles

Arquitectura: BiLSTM hidden=200, dropout=0.3, CRF de salida
Word embeddings: 300d (FastText)
CharCNN: 30d char emb -> Conv1d(filters=50, kernel=3) -> max-pool
Batch size: 8, épocas: 10
Compañeros: Laura Valero, Cristian Valencia
Taller PLN — Maestría Univalle