app.py

import torch
import gradio as gr
import numpy as np
from config import MLP

# Load model
model = MLP()
model.load_state_dict(torch.load("pytorch_model.pth", map_location=torch.device("cpu")))
model.eval()

# Example class names (you can change this)
class_names = [f"Class {i}" for i in range(8)]

# Prediction function
def predict(input_vector):
    input_array = np.array(input_vector).astype(np.float32)
    if len(input_array) != 1000:
        return "Error: Input must be 1000 numbers"
    tensor = torch.tensor(input_array).unsqueeze(0)
    with torch.no_grad():
        output = model(tensor)
        probs = torch.nn.functional.softmax(output[0], dim=0)
    return {class_names[i]: float(probs[i]) for i in range(8)}

# Gradio interface
demo = gr.Interface(
    fn=predict,
    inputs=gr.Textbox(lines=5, placeholder="Enter 1000 comma-separated numbers..."),
    outputs=gr.Label(num_top_classes=3),
    title="MLP Vector Classifier"
)

demo.launch()