modernbert-large-phishing

import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification

repo_id = "mikaelnurminen/modernbert-large-phishing"

tokenizer = AutoTokenizer.from_pretrained(repo_id, use_fast=True)
model = AutoModelForSequenceClassification.from_pretrained(
    repo_id,
    attn_implementation="eager",
).eval()

text = "EMAIL: Please verify your account details"
enc = tokenizer(text, return_tensors="pt", truncation=True, max_length=2048)
with torch.no_grad():
    logits = model(**enc).logits
prob_phishing = torch.softmax(logits.float(), dim=-1)[0, 1].item()
label = "phishing" if prob_phishing >= 0.47 else "benign"
print({"prob_phishing": prob_phishing, "label": label})