app.py

import gradio as gr
from transformers import pipeline

MODEL_ID = "pokwir/Bert_sentiment_classifier"
LABEL_ORDER = ["Neutral", "Positive", "Negative"]
ID2LABEL = {"LABEL_0": "Neutral", "LABEL_1": "Positive", "LABEL_2": "Negative"}

clf = pipeline(
    "text-classification",
    model=MODEL_ID,
    tokenizer=MODEL_ID,
    top_k=None  # replaces deprecated return_all_scores=True
)


def predict(text: str):
    text = (text or "").strip()
    if not text:
        return {lab: 0.0 for lab in LABEL_ORDER}, {}

    scores = clf(text)[0]  # list of dicts
    ordered = {lab: 0.0 for lab in LABEL_ORDER}

    for d in scores:
        lab = ID2LABEL.get(d["label"], d["label"])
        if lab in ordered:
            ordered[lab] = float(d["score"])

    return ordered, ordered

demo = gr.Interface(
    fn=predict,
    inputs=gr.Textbox(lines=3, placeholder="Type a sentence..."),
    outputs=[gr.Label(label="Scores"), gr.JSON(label="Scores (JSON)")],
    title="BERT Sentiment Classifier (3-class)",
    description=f"Model: {MODEL_ID}",
    examples=[
        ["I had surgery last month. and I was very impressed with the quality of service from the moment I got in till I left. Also I like to mention the nurses they were out standing"],
        ["I received the update and will review it later this week."],
        ["Dirty. Generally poor attitude among the nurses, even the good know the place sucks. When patients are crying for help nurse should not be busy watching Tik-Tok. Too many mistakes made too often. Teaching nurses instructing student nurse procedures incorrectly. Yes, it is bad."]
    ],
)

if __name__ == "__main__":
    demo.launch(ssr_mode=False)