import torch

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("davebulaval/MeaningBERT")
scorer = AutoModelForSequenceClassification.from_pretrained("davebulaval/MeaningBERT")
scorer.eval()

documents = [
    "He wanted to make them pay.",
    "This sandwich looks delicious.",
    "He wants to eat.",
]
simplifications = [
    "He wanted to make them pay.",
    "This sandwich looks delicious.",
    "Whatever, whenever, this is a sentence.",
]

# We tokenize the text as a pair and return Pytorch Tensors
tokenize_text = tokenizer(
    documents, simplifications, truncation=True, padding=True, return_tensors="pt"
)

with torch.no_grad():
    # We process the text
    scores = scorer(**tokenize_text)

print(scores.logits.tolist())