GPU.py

import os
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, BatchNormalization
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint, TensorBoard, ReduceLROnPlateau
import numpy as np
import datetime

# **Deteksi GPU & Konfigurasi TensorFlow**
gpus = tf.config.list_physical_devices('GPU')
if gpus:
    print(f"[+] GPU Ditemukan: {gpus}")
    os.environ["CUDA_VISIBLE_DEVICES"] = "0"
else:
    print("[!] Tidak ada GPU, menggunakan CPU...")
    os.environ["CUDA_VISIBLE_DEVICES"] = "-1"

# **Dataset Generator (Pastikan Shape Sesuai dengan Quze)**
def generate_data():
    """Membuat dataset dengan distribusi yang sesuai dengan input Quze (15 fitur)."""
    np.random.seed(42)
    X = np.random.normal(loc=0.0, scale=1.0, size=(10000, 15))  # **15 fitur**
    y = np.random.randint(0, 2, 10000)  # Label biner
    return X, y

# **Model AI Sesuai dengan Input Quze**
def create_model(input_shape):
    """Membangun model AI dengan arsitektur yang optimal & cocok dengan input Quze."""
    model = Sequential([
        Dense(256, input_shape=(input_shape,), activation='relu'),
        BatchNormalization(),
        Dropout(0.4),
        Dense(128, activation='relu'),
        BatchNormalization(),
        Dropout(0.3),
        Dense(64, activation='relu'),
        Dense(1, activation='sigmoid')
    ])
    model.compile(optimizer=Adam(learning_rate=0.001), 
                  loss='binary_crossentropy', 
                  metrics=['accuracy'])
    return model

# **Latih & Simpan Model (Pastikan Model Bisa Di-load di Quze)**
def train_and_save_models():
    X, y = generate_data()
    model = create_model(X.shape[1])

    print("[*] Mulai pelatihan model...")

    # **Callback untuk optimasi training**
    checkpoint_path = "best_model_v5.h5"
    log_dir = f"logs/fit/{datetime.datetime.now().strftime('%Y%m%d-%H%M%S')}"

    callbacks = [
        EarlyStopping(monitor='val_loss', patience=5, restore_best_weights=True),
        ModelCheckpoint(filepath=checkpoint_path, monitor='val_accuracy', save_best_only=True, mode='max'),
        ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=3, min_lr=1e-6, verbose=1),
        TensorBoard(log_dir=log_dir)
    ]

    # **Latih model dengan validasi**
    model.fit(X, y, epochs=100, batch_size=64, validation_split=0.2, callbacks=callbacks)

    # **Cek & Validasi Model Sebelum Disimpan**
    print("[*] Validasi model dengan input test...")
    test_input = np.random.rand(1, 15)  # **Pastikan input sesuai dengan model**
    try:
        test_output = model.predict(test_input)
        print(f"[+] Model valid, contoh output: {test_output}")
    except Exception as e:
        print(f"[!] ERROR: Model gagal dipakai untuk inferensi: {e}")
        return

    # **Simpan Model dengan Nama yang Konsisten**
    filenames = [
        "ml_model_V5.h5",
        "ml_model_v5.h5",
        "ml_Model_V5.h5",
        "ML_MODEL_V5.h5",
        "ml_model_v5.keras",
        "ml_model_v5_backup.h5"
    ]
    for name in filenames:
        if name.endswith(".h5"):
            model.save(name, save_format="h5")
        else:
            model.save(name)  # Format default Keras
        print(f"[+] Model berhasil disimpan sebagai {name}")

train_and_save_models()