app.py

# import gradio as gr
# from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline

# # Load model and tokenizer
# model_name = "chinesemusk/Deon_emotion-model-full"
# tokenizer_name = "chinesemusk/Deon-emotion-model-full"
# tokenizer = AutoTokenizer.from_pretrained(tokenizer_name)
# model = AutoModelForSequenceClassification.from_pretrained(model_name)

# # Map label IDs to readable names
# #label_names = [label for _, label in sorted(id2label.items(), key=lambda x: int(x[0]))]

# # Map label IDs to readable names
# id2label = model.config.id2label
# label_names = [id2label.get(str(i), id2label.get(i)) for i in range(len(id2label))]


# # Create inference pipeline
# classifier = pipeline("text-classification", model=model, tokenizer=tokenizer, top_k=None)

# def classify_emotion(text):
#     predictions = classifier(text)[0]
#     # Sort by score descending
#     sorted_preds = sorted(predictions, key=lambda x: x['score'], reverse=True)
#     # Convert label IDs to names
#     result = {id2label[x['label'].split('_')[-1]]: round(x['score'], 4) for x in sorted_preds}
#     return result

# examples = [
#     "I just got a new job and I'm so excited!",
#     "I'm feeling really down today.",
#     "I love spending time with my family.",
#     "That movie was terrifying!",
#     "I can't believe I won!",
#     "He really pissed me off with his attitude."
# ]

# # Launch Gradio interface
# gr.Interface(
#     fn=classify_emotion,
#     inputs=gr.Textbox(lines=2, placeholder="Enter a sentence to analyze emotion..."),
#     outputs=gr.Label(num_top_classes=6),
#     examples=examples,
#     title="Deon Emotion Classifier",
#     description="Classify text into one of 6 emotions: joy, sadness, anger, fear, love, and surprise. Powered by DistilBERT fine-tuned on emotion data.",
#     theme="default"
# ).launch() 
import gradio as gr
from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

# Load once globally
model_name = "chinesemusk/Deon_emotion-model-full"
tokenizer_name = "chinesemusk/Deon-emotion-model-full"
tokenizer = AutoTokenizer.from_pretrained(tokenizer_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Labels
labels = ["sadness", "joy", "love", "anger", "fear", "surprise"]

def classify_emotion(text):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
    with torch.no_grad():
        logits = model(**inputs).logits
    probs = torch.nn.functional.softmax(logits, dim=1)[0]
    return {labels[i]: float(probs[i]) for i in range(len(labels))}

# Gradio Interface
demo = gr.Interface(
    fn=classify_emotion,
    inputs=gr.Textbox(placeholder="Enter a sentence...", lines=2),
    outputs=gr.Label(num_top_classes=6),
    examples=[
        ["I feel amazing today!"],
        ["Why does this always happen to me?"],
        ["I'm terrified of what’s coming"],
        ["I’m deeply in love with her"],
        ["Nothing makes sense anymore"],
    ],
    title="Deon Emotion Detector",
    description="A fine-tuned DistilBERT model to classify text into six emotions: sadness, joy, love, anger, fear, and surprise.",
    allow_flagging="never"
)

demo.launch()