Japanese Natural Language Inference Model

This model was trained using SentenceTransformers Cross-Encoder class, gradient accumulation PR , and the code from CyberAgentAILab/japanese-nli-model .

Training Data

The model was trained on the JGLUE-JNLI and JSICK datasets. For a given sentence pair, it will output three scores corresponding to the labels: contradiction, entailment, neutral.

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained('cyberagent/xlm-roberta-large-jnli-jsick')
model = AutoModelForSequenceClassification.from_pretrained('cyberagent/xlm-roberta-large-jnli-jsick')
features = tokenizer(["子供が走っている猫を見ている", "猫が走っている"], ["猫が走っている", "子供が走っている"], padding=True, truncation=True, return_tensors="pt")
model.eval()
with torch.no_grad():
    scores = model(**features).logits
    label_mapping = ['contradiction', 'entailment', 'neutral']
    labels = [label_mapping[score_max] for score_max in scores.argmax(dim=1)]
    print(labels)