app.py

import gradio as gr
import numpy as np
import torch

from utils.data_processing import get_data
from modules.tgn import TGN

# Load data
node_features, edge_features, full_data, train_data, val_data, test_data, \
    new_node_val_data, new_node_test_data = get_data('wikipedia')

# Initialize model (must match training args)
tgn = TGN(
    raw_message_dim=edge_features.shape[1],
    memory_dim=172,
    message_dim=100,
    memory_update_at_end=False,
    embedding_module_type='graph_attention',
    message_function='identity',
    memory_updater_type='gru',
    n_heads=2,
    dropout=0.1,
    n_layers=1,
    n_neighbors=10,
    aggregator_type='last',
    edge_features=edge_features,
    node_features=node_features,
    device='cpu',
    use_memory=True,
)
tgn.load_state_dict(torch.load("tgn-attn-wikipedia.pth", map_location="cpu"))
tgn.eval()

def predict(u_id, i_id, timestamp):
    u = np.array([int(u_id)])
    i = np.array([int(i_id)])
    ts = np.array([float(timestamp)])
    prob = tgn.predict_edge_probabilities(u, i, ts)
    return f"Predicted interaction probability: {prob[0]:.4f}"

demo = gr.Interface(
    fn=predict,
    inputs=[
        gr.Number(label="Source Node ID"),
        gr.Number(label="Destination Node ID"),
        gr.Number(label="Timestamp"),
    ],
    outputs="text",
    title="🧠 TGN Playground (Wikipedia)",
    description="Enter node IDs and timestamp to predict future interaction probability using Temporal Graph Networks (TGN).",
)

demo.launch()