from tensorflow.keras.datasets import mnist
from tensorflow.keras.utils import to_categorical
import numpy as np


(x_train, y_train), (x_test, y_test) = mnist.load_data()

x_train = x_train.astype("float32") / 255.0
x_test = x_test.astype("float32") / 255.0

x_train = x_train[..., np.newaxis]
x_test = x_test[..., np.newaxis]

y_train = to_categorical(y_train, 10)
y_test = to_categorical(y_test, 10)

#Wycięcie 10 przykładów ze zbioru testowego
images_to_check = x_test[:10]
labels_to_check = y_test[:10]

x_test = x_test[10:]
y_test = y_test[10:]

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout, BatchNormalization
import time
start = time.time()

model = Sequential([
    #Pierwszy blok konwolucyjny
    Conv2D(32, (3, 3), activation='relu', padding='same', input_shape=(28, 28, 1)),
    BatchNormalization(),
    MaxPooling2D((2, 2)),
    Dropout(0.25),

    #Drugi blok konwolucyjny
    Conv2D(64, (3, 3), activation='relu', padding='same'),
    BatchNormalization(),
    MaxPooling2D((2, 2)),
    Dropout(0.25),

    #Blok gęsty
    Flatten(),
    Dense(128, activation='relu'),
    BatchNormalization(),
    Dropout(0.5),
    Dense(10, activation='softmax')
])

model.compile(optimizer="adam",
              loss="categorical_crossentropy",
              metrics=["accuracy"])


history = model.fit(x_train, y_train,
                    epochs=10,
                    batch_size=32,
                    validation_data=(x_test, y_test),
                    callbacks=[tensorboard_callback])


loss, accuracy = model.evaluate(x_test, y_test, verbose=0)
print(f"Strata: {loss:.4f}, Dokładność: {accuracy:.4f}")

end = time.time()  # czas końca

print(f"Czas wykonania: {end - start:.2f} sekundy")