DeBERTa-v3-base-mnli-fever-docnli-ling-2c

Clasificación simple cero disparo:

from transformers import pipeline
classifier = pipeline("zero-shot-classification", model="MoritzLaurer/DeBERTa-v3-base-mnli-fever-docnli-ling-2c")
sequence_to_classify = "Angela Merkel es una política en Alemania y líder de la CDU"
candidate_labels = ["política", "economía", "entretenimiento", "medio ambiente"]
output = classifier(sequence_to_classify, candidate_labels, multi_label=False)
print(output)

Caso de uso de NLI:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")

model_name = "MoritzLaurer/DeBERTa-v3-base-mnli-fever-docnli-ling-2c"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

premise = "Al principio pensé que me gustó la película, pero después de reconsiderarlo, me resultó decepcionante."
hypothesis = "La película era buena."

input = tokenizer(premise, hypothesis, truncation=True, return_tensors="pt")
output = model(input["input_ids"].to(device)) # dispositivo = "cuda:0" o "cpu"
prediction = torch.softmax(output["logits"][0], -1).tolist()
label_names = ["implica", "no_implica"]
prediction = {name: round(float(pred) * 100, 1) for pred, name in zip(prediction, label_names)}
print(prediction)