app.py

import torch
import numpy as np
import pandas as pd
import gradio as gr
import joblib
from model import LSTMModel

# Constants
input_size = 20

# Load model and scalers
model = LSTMModel(input_size=input_size)
model.load_state_dict(torch.load("model.pth", map_location="cpu"))
model.eval()

input_scaler = joblib.load("input_scaler.pkl")
target_scaler = joblib.load("target_scaler.pkl")

# Prediction logic
def predict_from_array(array_2d):
    if array_2d.shape != (10, input_size):
        return f"❌ Expected shape (10, {input_size}), got {array_2d.shape}"
    try:
        scaled = input_scaler.transform(array_2d)
        tensor = torch.tensor(scaled, dtype=torch.float32).unsqueeze(0)
        with torch.no_grad():
            pred_scaled = model(tensor).item()
        pred_real = target_scaler.inverse_transform([[pred_scaled]])[0][0]
        return f"✅ Predicted Plus/Minus: {pred_real:.2f}"
    except Exception as e:
        return f"❌ Error during prediction: {e}"

# CSV Upload Handler
def predict_from_csv(file):
    try:
        df = pd.read_csv(file.name)
        if df.shape != (10, input_size):
            return f"❌ CSV must have shape (10, {input_size}). Got {df.shape}"
        return predict_from_array(df.to_numpy(dtype=np.float32))
    except Exception as e:
        return f"❌ Error reading CSV: {e}"

# Generate random input
def generate_random_input():
    return pd.DataFrame(np.random.uniform(0, 100, size=(10, input_size)))

def predict_from_table(df):
    try:
        if df.shape != (10, input_size):
            return f"❌ Table must have shape (10, {input_size})"
        return predict_from_array(df.to_numpy(dtype=np.float32))
    except Exception as e:
        return f"❌ Error: {e}"

# Build UI
with gr.Blocks() as app:
    gr.Markdown("## 🏀 NBA Plus/Minus Predictor")

    with gr.Tab("🎲 Generate Random Input"):
        with gr.Row():
            btn = gr.Button("Generate Random 10x20 Sample")
            predict_btn = gr.Button("Predict")

        input_table = gr.Dataframe(
            headers=[f"F{i+1}" for i in range(input_size)],
            row_count=10, col_count=input_size,
            interactive=True,
            label="Generated Features"
        )
        output_text = gr.Textbox(label="Prediction")

        btn.click(fn=generate_random_input, outputs=input_table)
        predict_btn.click(fn=predict_from_table, inputs=input_table, outputs=output_text)

    with gr.Tab("📂 Upload CSV"):
        with gr.Row():
            file_input = gr.File(label="Upload CSV with shape (10×20)")
            predict_btn = gr.Button("Predict")
        
        file_output = gr.Textbox(label="Prediction")

        predict_btn.click(fn=predict_from_csv, inputs=file_input, outputs=file_output)

app.launch(share=True, show_api=True, show_error=True)