from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_id = "darwinkernelpanic/ai-detector-pgx"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSequenceClassification.from_pretrained(model_id)

text = "The mitochondria is the powerhouse of the cell..."
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)

with torch.no_grad():
    outputs = model(**inputs)
    probs = torch.softmax(outputs.logits, dim=1)
    ai_prob = probs[0][1].item()

print(f"AI Probability: {ai_prob:.2%}")

JavaScript (ONNX)

javascript

import * as ort from 'onnxruntime-web';

const session = await ort.InferenceSession.create('model.onnx');
// Tokenize with @xenova/transformers, then run inference
const results = await session.run({ input_ids, attention_mask });
const logits = results.logits.data;
const aiProb = Math.exp(logits[1]) / (Math.exp(logits[0]) + Math.exp(logits[1]));

Model Details

Base: prajjwal1/bert-tiny (4.4M params)
Classes: human (0), ai (1)
Sequence length: 512 tokens
ONNX size: 255MB

Limitations

Trained on academic essays — may not generalize to all text types.