app.py

import gradio as gr
from optimum.onnxruntime import ORTModelForSequenceClassification
from transformers import pipeline, AutoTokenizer

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained("iam-tsr/finetuned-distilbert-employ-feedback")

# Load Quantized ONNX model
onnx_filename = "model_qint8.onnx"
model = ORTModelForSequenceClassification.from_pretrained("iam-tsr/finetuned-distilbert-employ-feedback", file_name=onnx_filename)

def pred(model, tokenizer, text):
    
    pipe = pipeline(task="text-classification", model=model, tokenizer=tokenizer, device="cpu")
    return pipe(text)[0]['label']

demo = gr.Interface(
    fn=lambda text: pred(model, tokenizer, text),
    inputs=["text"],
    outputs=["text"],
    api_name="predict"
)

demo.launch(share=True)