app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Model name (you can swap this for another emotion model if you like)
model_name = "j-hartmann/emotion-english-distilroberta-base"
#minoosh/finetuned_bert-base-on-IEMOCAP_1

# Device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Load tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name).to(device)
model.eval()

# Prediction function
def predict_emotion(text: str):
    # Handle empty input
    if not text or not text.strip():
        return {"Error": "Please enter some text."}

    # Tokenize
    inputs = tokenizer(
        text,
        return_tensors="pt",
        truncation=True,
        padding=True,
        max_length=256,  # you can adjust this if needed
    ).to(device)

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

    # Map id -> label using model config
    id2label = model.config.id2label
    scores = {id2label[i]: float(probs[i]) for i in range(len(probs))}

    # Sort by highest probability first (optional but nice in the UI)
    scores = dict(sorted(scores.items(), key=lambda x: x[1], reverse=True))
    return scores

# Gradio interface
demo = gr.Interface(
    fn=predict_emotion,
    inputs=gr.Textbox(lines=4, label="Enter text"),
    outputs=gr.Label(label="Emotion Probabilities"),
    title="Emotion Classifier",
    description="Enter a sentence and see the predicted emotion distribution.",
    flagging_mode="never",
)

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