app.py

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

tokenizer = AutoTokenizer.from_pretrained("akhooli/mistral-7B-llm")
model = AutoModelForSequenceClassification.from_pretrained("akhooli/mistral-7B-llm")

def predict(image_path):
    # Load and preprocess the image
    with open(image_path, "rb") as f:
        image_bytes = f.read()

    # Tokenize and predict
    inputs = tokenizer(image_bytes, return_tensors="pt", padding=True, truncation=True)
    outputs = model(**inputs)
    predicted_class_idx = torch.argmax(outputs.logits)
    
    # In this example, we are assuming the labels are ['pizza', 'burger', 'sandwich']
    labels = ['pizza', 'burger', 'sandwich']
    predicted_label = labels[predicted_class_idx]

    return predicted_label

gr.Interface(
    predict,
    inputs=gr.Image(label="Upload junk food (sandwich, pizza, burger) candidate", type="file"),
    outputs=gr.Label(num_top_classes=3),
    title="Pizza, Burger, or Sandwich?",
).launch()