Naver Project (fashion_MNIST-keras)

import tensorflow as tf
from tensorflow.keras import layers
from tensorflow.keras import callbacks

from sklearn.model_selection import train_test_split

import numpy as np
import matplotlib.pyplot as plt

tf.__version__

#2.15.0

데이터 로드

• TF에서 제공하는 데이터셋을 load해 간단한 전처리 진행

# Load training and eval data from tf.keras
(train_data, train_labels), (test_data, test_labels) = \
    tf.keras.datasets.fashion_mnist.load_data()

print(train_data.shape, train_labels.shape)

#(60000, 28, 28) (60000,)

train_data, valid_data, train_labels, valid_labels = \
    train_test_split(train_data, train_labels, test_size=0.1, shuffle=True)

print(train_data.shape, train_labels.shape)

#(54000, 28, 28) (54000,)

print(valid_data.shape, valid_labels.shape)

#(6000, 28, 28) (6000,)

index = 5901
print("label = {}".format(valid_labels[index]))
plt.imshow(valid_data[index].reshape(28, 28))
plt.colorbar()
#plt.gca().grid(False)
plt.show()

train_data = train_data / 255.
train_data = train_data.reshape(-1, 28 * 28)
train_data = train_data.astype(np.float32)
train_labels = train_labels.astype(np.int32)

test_data = test_data / 255.
test_data = test_data.reshape(-1, 28 * 28)
test_data = test_data.astype(np.float32)
test_labels = test_labels.astype(np.int32)

valid_data = valid_data / 255.
valid_data = valid_data.reshape(-1, 28 * 28)
valid_data = valid_data.astype(np.float32)
valid_labels = valid_labels.astype(np.int32)


print(train_data.shape, train_labels.shape)
print(test_data.shape, test_labels.shape)
print(valid_data.shape, valid_labels.shape)

#(54000, 784) (54000,)
#(10000, 784) (10000,)
#(6000, 784) (6000,)

학습에 사용할 tf.data.Dataset 구성

• 학습에 잘 적용할 수 있도록 label 처리
• 데이터셋 구성

def one_hot_label(image, label):
  label = tf.one_hot(label, depth=10)
  return image, label

batch_size = 32

# for train
train_dataset = tf.data.Dataset.from_tensor_slices((train_data, train_labels))
train_dataset = train_dataset.shuffle(buffer_size=10000)
train_dataset = train_dataset.map(one_hot_label)
train_dataset = train_dataset.repeat().batch(batch_size=batch_size)
print(train_dataset)

# for test
test_dataset = tf.data.Dataset.from_tensor_slices((test_data, test_labels))
test_dataset = test_dataset.map(one_hot_label)
test_dataset = test_dataset.batch(batch_size=batch_size)
print(test_dataset)

# for test
valid_dataset = tf.data.Dataset.from_tensor_slices((valid_data, valid_labels))
valid_dataset = valid_dataset.map(one_hot_label)
valid_dataset = valid_dataset.batch(batch_size=batch_size)
print(valid_dataset)

#<_BatchDataset element_spec=(TensorSpec(shape=(None, 784), dtype=tf.float32, name=None), TensorSpec(shape=(None, 10), dtype=tf.float32, name=None))>
#<_BatchDataset element_spec=(TensorSpec(shape=(None, 784), dtype=tf.float32, name=None), TensorSpec(shape=(None, 10), dtype=tf.float32, name=None))>
#<_BatchDataset element_spec=(TensorSpec(shape=(None, 784), dtype=tf.float32, name=None), TensorSpec(shape=(None, 10), dtype=tf.float32, name=None))>

모델 구성

# Dense 64 relu
# Dense 32 relu
# Output Layer
# 28 * 28
model = tf.keras.Sequential([
    layers.Dense(64, kernel_initializer=tf.keras.initializers.HeUniform(),
                 kernel_regularizer=tf.keras.regularizers.L2(0.0001)),
    layers.BatchNormalization(),
    layers.Activation("relu"),

    layers.Dropout(0.5),

    layers.Dense(32, kernel_initializer=tf.keras.initializers.HeUniform(),
                 kernel_regularizer=tf.keras.regularizers.L2(0.0001)),
    layers.BatchNormalization(),
    layers.Activation("relu"),

    layers.Dropout(0.5),

    layers.Dense(16, kernel_initializer=tf.keras.initializers.HeUniform(),
                 kernel_regularizer=tf.keras.regularizers.L2(0.0001)),
    layers.BatchNormalization(),
    layers.Activation("relu"),

    layers.Dense(10),
])

model.compile(optimizer=tf.keras.optimizers.Adam(1e-4),
              loss=tf.keras.losses.CategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

# without training, just inference a model in eager execution:
predictions = model(train_data[0:1], training=False)
print("Predictions: ", predictions.numpy())

#Predictions:  [[-1.0235932  -0.24154036 -0.50594735  0.3415959   #0.8167548   0.4199386
#   0.4468755   0.7400886   0.15017883 -0.7716306 ]]

model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense (Dense)               (1, 64)                   50240     
                                                                 
 batch_normalization (Batch  (1, 64)                   256       
 Normalization)                                                  
                                                                 
 activation (Activation)     (1, 64)                   0         
                                                                 
 dropout (Dropout)           (1, 64)                   0         
                                                                 
 dense_1 (Dense)             (1, 32)                   2080      
                                                                 
 batch_normalization_1 (Bat  (1, 32)                   128       
 chNormalization)                                                
                                                                 
 activation_1 (Activation)   (1, 32)                   0         
                                                                 
 dropout_1 (Dropout)         (1, 32)                   0         
                                                                 
 dense_2 (Dense)             (1, 16)                   528       
                                                                 
 batch_normalization_2 (Bat  (1, 16)                   64        
 chNormalization)                                                
                                                                 
 activation_2 (Activation)   (1, 16)                   0         
                                                                 
 dense_3 (Dense)             (1, 10)                   170       
                                                                 
=================================================================
Total params: 53466 (208.85 KB)
Trainable params: 53242 (207.98 KB)
Non-trainable params: 224 (896.00 Byte)
_________________________________________________________________

모델 학습 구성

check_point_cb = callbacks.ModelCheckpoint('fashion_mnist_model.h5',
                                           save_best_only=True,
                                           verbose=1)
early_stopping_cb = callbacks.EarlyStopping(patience=10,
                                            monitor='val_loss',
                                            restore_best_weights=True,
                                            verbose=1)

max_epochs = 100
# using `tf.data.Dataset`
history = model.fit(train_dataset,
                    epochs=max_epochs,
                    steps_per_epoch=len(train_data) // batch_size,
                    validation_data=valid_dataset,
                    validation_steps=len(valid_data) // batch_size,
                    callbacks=[check_point_cb, early_stopping_cb]
                   )

Epoch 1/100
1680/1687 [============================>.] - ETA: 0s - loss: 1.8866 - accuracy: 0.3592
Epoch 1: val_loss improved from inf to 1.35230, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 27s 11ms/step - loss: 1.8853 - accuracy: 0.3600 - val_loss: 1.3523 - val_accuracy: 0.7495
Epoch 2/100
  18/1687 [..............................] - ETA: 10s - loss: 1.6272 - accuracy: 0.4792/usr/local/lib/python3.10/dist-packages/keras/src/engine/training.py:3103: UserWarning: You are saving your model as an HDF5 file via `model.save()`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')`.
  saving_api.save_model(
1685/1687 [============================>.] - ETA: 0s - loss: 1.4218 - accuracy: 0.5672
Epoch 2: val_loss improved from 1.35230 to 0.95672, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 12s 7ms/step - loss: 1.4215 - accuracy: 0.5672 - val_loss: 0.9567 - val_accuracy: 0.7871
Epoch 3/100
1687/1687 [==============================] - ETA: 0s - loss: 1.1770 - accuracy: 0.6316
Epoch 3: val_loss improved from 0.95672 to 0.75706, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 7ms/step - loss: 1.1770 - accuracy: 0.6316 - val_loss: 0.7571 - val_accuracy: 0.7968
Epoch 4/100
1684/1687 [============================>.] - ETA: 0s - loss: 1.0440 - accuracy: 0.6617
Epoch 4: val_loss improved from 0.75706 to 0.66723, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 7ms/step - loss: 1.0442 - accuracy: 0.6617 - val_loss: 0.6672 - val_accuracy: 0.8107
Epoch 5/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.9719 - accuracy: 0.6803
Epoch 5: val_loss improved from 0.66723 to 0.59835, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.9722 - accuracy: 0.6802 - val_loss: 0.5983 - val_accuracy: 0.8175
Epoch 6/100
1687/1687 [==============================] - ETA: 0s - loss: 0.9129 - accuracy: 0.6979
Epoch 6: val_loss improved from 0.59835 to 0.56844, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 12s 7ms/step - loss: 0.9129 - accuracy: 0.6979 - val_loss: 0.5684 - val_accuracy: 0.8284
Epoch 7/100
1683/1687 [============================>.] - ETA: 0s - loss: 0.8799 - accuracy: 0.7058
Epoch 7: val_loss improved from 0.56844 to 0.54611, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.8803 - accuracy: 0.7055 - val_loss: 0.5461 - val_accuracy: 0.8319
Epoch 8/100
1687/1687 [==============================] - ETA: 0s - loss: 0.8536 - accuracy: 0.7158
Epoch 8: val_loss improved from 0.54611 to 0.52750, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.8536 - accuracy: 0.7158 - val_loss: 0.5275 - val_accuracy: 0.8334
Epoch 9/100
1687/1687 [==============================] - ETA: 0s - loss: 0.8268 - accuracy: 0.7255
Epoch 9: val_loss improved from 0.52750 to 0.51081, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 14s 8ms/step - loss: 0.8268 - accuracy: 0.7255 - val_loss: 0.5108 - val_accuracy: 0.8386
Epoch 10/100
1683/1687 [============================>.] - ETA: 0s - loss: 0.8107 - accuracy: 0.7318
Epoch 10: val_loss improved from 0.51081 to 0.49830, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.8103 - accuracy: 0.7320 - val_loss: 0.4983 - val_accuracy: 0.8379
Epoch 11/100
1681/1687 [============================>.] - ETA: 0s - loss: 0.7928 - accuracy: 0.7361
Epoch 11: val_loss improved from 0.49830 to 0.49690, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 14s 8ms/step - loss: 0.7923 - accuracy: 0.7363 - val_loss: 0.4969 - val_accuracy: 0.8384
Epoch 12/100
1685/1687 [============================>.] - ETA: 0s - loss: 0.7868 - accuracy: 0.7385
Epoch 12: val_loss improved from 0.49690 to 0.48387, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.7867 - accuracy: 0.7385 - val_loss: 0.4839 - val_accuracy: 0.8443
Epoch 13/100
1687/1687 [==============================] - ETA: 0s - loss: 0.7689 - accuracy: 0.7449
Epoch 13: val_loss improved from 0.48387 to 0.47823, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 14s 8ms/step - loss: 0.7689 - accuracy: 0.7449 - val_loss: 0.4782 - val_accuracy: 0.8421
Epoch 14/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.7628 - accuracy: 0.7463
Epoch 14: val_loss improved from 0.47823 to 0.47363, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.7627 - accuracy: 0.7464 - val_loss: 0.4736 - val_accuracy: 0.8436
Epoch 15/100
1683/1687 [============================>.] - ETA: 0s - loss: 0.7465 - accuracy: 0.7528
Epoch 15: val_loss improved from 0.47363 to 0.46710, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.7466 - accuracy: 0.7528 - val_loss: 0.4671 - val_accuracy: 0.8446
Epoch 16/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.7388 - accuracy: 0.7560
Epoch 16: val_loss improved from 0.46710 to 0.46406, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.7385 - accuracy: 0.7562 - val_loss: 0.4641 - val_accuracy: 0.8486
Epoch 17/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.7294 - accuracy: 0.7589
Epoch 17: val_loss improved from 0.46406 to 0.45721, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.7295 - accuracy: 0.7589 - val_loss: 0.4572 - val_accuracy: 0.8491
Epoch 18/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.7235 - accuracy: 0.7599
Epoch 18: val_loss improved from 0.45721 to 0.45413, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.7236 - accuracy: 0.7599 - val_loss: 0.4541 - val_accuracy: 0.8508
Epoch 19/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.7156 - accuracy: 0.7651
Epoch 19: val_loss improved from 0.45413 to 0.45155, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 18s 10ms/step - loss: 0.7154 - accuracy: 0.7651 - val_loss: 0.4515 - val_accuracy: 0.8498
Epoch 20/100
1687/1687 [==============================] - ETA: 0s - loss: 0.7162 - accuracy: 0.7643
Epoch 20: val_loss did not improve from 0.45155
1687/1687 [==============================] - 14s 8ms/step - loss: 0.7162 - accuracy: 0.7643 - val_loss: 0.4586 - val_accuracy: 0.8459
Epoch 21/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.7086 - accuracy: 0.7666
Epoch 21: val_loss improved from 0.45155 to 0.44700, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.7085 - accuracy: 0.7667 - val_loss: 0.4470 - val_accuracy: 0.8483
Epoch 22/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.7045 - accuracy: 0.7673
Epoch 22: val_loss did not improve from 0.44700
1687/1687 [==============================] - 13s 8ms/step - loss: 0.7046 - accuracy: 0.7672 - val_loss: 0.4487 - val_accuracy: 0.8489
Epoch 23/100
1681/1687 [============================>.] - ETA: 0s - loss: 0.6985 - accuracy: 0.7701
Epoch 23: val_loss improved from 0.44700 to 0.44265, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6988 - accuracy: 0.7699 - val_loss: 0.4426 - val_accuracy: 0.8531
Epoch 24/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.6889 - accuracy: 0.7735
Epoch 24: val_loss improved from 0.44265 to 0.43872, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6886 - accuracy: 0.7736 - val_loss: 0.4387 - val_accuracy: 0.8519
Epoch 25/100
1681/1687 [============================>.] - ETA: 0s - loss: 0.6864 - accuracy: 0.7735
Epoch 25: val_loss improved from 0.43872 to 0.43230, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6863 - accuracy: 0.7734 - val_loss: 0.4323 - val_accuracy: 0.8566
Epoch 26/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.6871 - accuracy: 0.7737
Epoch 26: val_loss did not improve from 0.43230
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6868 - accuracy: 0.7739 - val_loss: 0.4404 - val_accuracy: 0.8506
Epoch 27/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6769 - accuracy: 0.7776
Epoch 27: val_loss did not improve from 0.43230
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6768 - accuracy: 0.7776 - val_loss: 0.4374 - val_accuracy: 0.8529
Epoch 28/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6840 - accuracy: 0.7749
Epoch 28: val_loss did not improve from 0.43230
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6840 - accuracy: 0.7749 - val_loss: 0.4345 - val_accuracy: 0.8549
Epoch 29/100
1683/1687 [============================>.] - ETA: 0s - loss: 0.6788 - accuracy: 0.7788
Epoch 29: val_loss did not improve from 0.43230
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6785 - accuracy: 0.7790 - val_loss: 0.4328 - val_accuracy: 0.8514
Epoch 30/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6714 - accuracy: 0.7797
Epoch 30: val_loss did not improve from 0.43230
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6715 - accuracy: 0.7796 - val_loss: 0.4349 - val_accuracy: 0.8561
Epoch 31/100
1685/1687 [============================>.] - ETA: 0s - loss: 0.6676 - accuracy: 0.7800
Epoch 31: val_loss improved from 0.43230 to 0.43181, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 15s 9ms/step - loss: 0.6678 - accuracy: 0.7799 - val_loss: 0.4318 - val_accuracy: 0.8544
Epoch 32/100
1683/1687 [============================>.] - ETA: 0s - loss: 0.6640 - accuracy: 0.7817
Epoch 32: val_loss improved from 0.43181 to 0.42382, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 14s 9ms/step - loss: 0.6640 - accuracy: 0.7816 - val_loss: 0.4238 - val_accuracy: 0.8573
Epoch 33/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6670 - accuracy: 0.7816
Epoch 33: val_loss did not improve from 0.42382
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6669 - accuracy: 0.7817 - val_loss: 0.4279 - val_accuracy: 0.8539
Epoch 34/100
1681/1687 [============================>.] - ETA: 0s - loss: 0.6595 - accuracy: 0.7837
Epoch 34: val_loss improved from 0.42382 to 0.42215, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6596 - accuracy: 0.7836 - val_loss: 0.4222 - val_accuracy: 0.8573
Epoch 35/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6551 - accuracy: 0.7850
Epoch 35: val_loss did not improve from 0.42215
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6553 - accuracy: 0.7850 - val_loss: 0.4240 - val_accuracy: 0.8583
Epoch 36/100
1681/1687 [============================>.] - ETA: 0s - loss: 0.6541 - accuracy: 0.7870
Epoch 36: val_loss did not improve from 0.42215
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6544 - accuracy: 0.7870 - val_loss: 0.4296 - val_accuracy: 0.8558
Epoch 37/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.6549 - accuracy: 0.7841
Epoch 37: val_loss improved from 0.42215 to 0.42087, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6547 - accuracy: 0.7842 - val_loss: 0.4209 - val_accuracy: 0.8595
Epoch 38/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.6474 - accuracy: 0.7894
Epoch 38: val_loss improved from 0.42087 to 0.41692, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6472 - accuracy: 0.7895 - val_loss: 0.4169 - val_accuracy: 0.8620
Epoch 39/100
1680/1687 [============================>.] - ETA: 0s - loss: 0.6471 - accuracy: 0.7869
Epoch 39: val_loss did not improve from 0.41692
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6467 - accuracy: 0.7870 - val_loss: 0.4232 - val_accuracy: 0.8610
Epoch 40/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.6437 - accuracy: 0.7886
Epoch 40: val_loss did not improve from 0.41692
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6435 - accuracy: 0.7887 - val_loss: 0.4208 - val_accuracy: 0.8591
Epoch 41/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6419 - accuracy: 0.7901
Epoch 41: val_loss did not improve from 0.41692
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6419 - accuracy: 0.7901 - val_loss: 0.4183 - val_accuracy: 0.8583
Epoch 42/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6399 - accuracy: 0.7906
Epoch 42: val_loss did not improve from 0.41692
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6399 - accuracy: 0.7905 - val_loss: 0.4245 - val_accuracy: 0.8570
Epoch 43/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.6456 - accuracy: 0.7901
Epoch 43: val_loss improved from 0.41692 to 0.41269, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6455 - accuracy: 0.7901 - val_loss: 0.4127 - val_accuracy: 0.8620
Epoch 44/100
1683/1687 [============================>.] - ETA: 0s - loss: 0.6362 - accuracy: 0.7916
Epoch 44: val_loss improved from 0.41269 to 0.41099, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6362 - accuracy: 0.7916 - val_loss: 0.4110 - val_accuracy: 0.8623
Epoch 45/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.6365 - accuracy: 0.7912
Epoch 45: val_loss improved from 0.41099 to 0.40983, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6365 - accuracy: 0.7912 - val_loss: 0.4098 - val_accuracy: 0.8665
Epoch 46/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6336 - accuracy: 0.7949
Epoch 46: val_loss improved from 0.40983 to 0.40962, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6335 - accuracy: 0.7950 - val_loss: 0.4096 - val_accuracy: 0.8608
Epoch 47/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.6354 - accuracy: 0.7922
Epoch 47: val_loss improved from 0.40962 to 0.40934, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6354 - accuracy: 0.7922 - val_loss: 0.4093 - val_accuracy: 0.8648
Epoch 48/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6326 - accuracy: 0.7918
Epoch 48: val_loss did not improve from 0.40934
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6326 - accuracy: 0.7918 - val_loss: 0.4107 - val_accuracy: 0.8638
Epoch 49/100
1685/1687 [============================>.] - ETA: 0s - loss: 0.6290 - accuracy: 0.7953
Epoch 49: val_loss improved from 0.40934 to 0.40505, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6288 - accuracy: 0.7954 - val_loss: 0.4051 - val_accuracy: 0.8651
Epoch 50/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6242 - accuracy: 0.7945
Epoch 50: val_loss did not improve from 0.40505
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6243 - accuracy: 0.7945 - val_loss: 0.4092 - val_accuracy: 0.8656
Epoch 51/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.6224 - accuracy: 0.7960
Epoch 51: val_loss did not improve from 0.40505
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6225 - accuracy: 0.7960 - val_loss: 0.4057 - val_accuracy: 0.8633
Epoch 52/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.6285 - accuracy: 0.7931
Epoch 52: val_loss did not improve from 0.40505
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6286 - accuracy: 0.7930 - val_loss: 0.4059 - val_accuracy: 0.8638
Epoch 53/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.6216 - accuracy: 0.7966
Epoch 53: val_loss did not improve from 0.40505
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6217 - accuracy: 0.7965 - val_loss: 0.4063 - val_accuracy: 0.8660
Epoch 54/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6181 - accuracy: 0.7976
Epoch 54: val_loss did not improve from 0.40505
1687/1687 [==============================] - 16s 9ms/step - loss: 0.6181 - accuracy: 0.7976 - val_loss: 0.4075 - val_accuracy: 0.8640
Epoch 55/100
1681/1687 [============================>.] - ETA: 0s - loss: 0.6198 - accuracy: 0.7964
Epoch 55: val_loss did not improve from 0.40505
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6199 - accuracy: 0.7965 - val_loss: 0.4104 - val_accuracy: 0.8623
Epoch 56/100
1685/1687 [============================>.] - ETA: 0s - loss: 0.6186 - accuracy: 0.7981
Epoch 56: val_loss did not improve from 0.40505
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6183 - accuracy: 0.7981 - val_loss: 0.4086 - val_accuracy: 0.8630
Epoch 57/100
1684/1687 [============================>.] - ETA: 0s - loss: 0.6181 - accuracy: 0.7968
Epoch 57: val_loss did not improve from 0.40505
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6180 - accuracy: 0.7969 - val_loss: 0.4087 - val_accuracy: 0.8643
Epoch 58/100
1683/1687 [============================>.] - ETA: 0s - loss: 0.6190 - accuracy: 0.7974
Epoch 58: val_loss improved from 0.40505 to 0.40263, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6192 - accuracy: 0.7972 - val_loss: 0.4026 - val_accuracy: 0.8628
Epoch 59/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6106 - accuracy: 0.7976
Epoch 59: val_loss did not improve from 0.40263
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6106 - accuracy: 0.7976 - val_loss: 0.4152 - val_accuracy: 0.8608
Epoch 60/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6179 - accuracy: 0.7963
Epoch 60: val_loss did not improve from 0.40263
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6179 - accuracy: 0.7963 - val_loss: 0.4053 - val_accuracy: 0.8661
Epoch 61/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.6140 - accuracy: 0.7976
Epoch 61: val_loss did not improve from 0.40263
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6140 - accuracy: 0.7977 - val_loss: 0.4048 - val_accuracy: 0.8641
Epoch 62/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6122 - accuracy: 0.7997
Epoch 62: val_loss did not improve from 0.40263
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6122 - accuracy: 0.7997 - val_loss: 0.4035 - val_accuracy: 0.8675
Epoch 63/100
1682/1687 [============================>.] - ETA: 0s - loss: 0.6078 - accuracy: 0.8007
Epoch 63: val_loss improved from 0.40263 to 0.39922, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6081 - accuracy: 0.8006 - val_loss: 0.3992 - val_accuracy: 0.8671
Epoch 64/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6084 - accuracy: 0.8007
Epoch 64: val_loss did not improve from 0.39922
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6084 - accuracy: 0.8007 - val_loss: 0.4019 - val_accuracy: 0.8646
Epoch 65/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6022 - accuracy: 0.8031
Epoch 65: val_loss did not improve from 0.39922
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6022 - accuracy: 0.8031 - val_loss: 0.4036 - val_accuracy: 0.8638
Epoch 66/100
1683/1687 [============================>.] - ETA: 0s - loss: 0.6036 - accuracy: 0.8015
Epoch 66: val_loss improved from 0.39922 to 0.39657, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6034 - accuracy: 0.8015 - val_loss: 0.3966 - val_accuracy: 0.8707
Epoch 67/100
1685/1687 [============================>.] - ETA: 0s - loss: 0.6051 - accuracy: 0.7997
Epoch 67: val_loss did not improve from 0.39657
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6048 - accuracy: 0.7998 - val_loss: 0.4001 - val_accuracy: 0.8670
Epoch 68/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6053 - accuracy: 0.8030
Epoch 68: val_loss did not improve from 0.39657
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6053 - accuracy: 0.8030 - val_loss: 0.4028 - val_accuracy: 0.8650
Epoch 69/100
1687/1687 [==============================] - ETA: 0s - loss: 0.6042 - accuracy: 0.8031
Epoch 69: val_loss did not improve from 0.39657
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6042 - accuracy: 0.8031 - val_loss: 0.4035 - val_accuracy: 0.8661
Epoch 70/100
1685/1687 [============================>.] - ETA: 0s - loss: 0.6072 - accuracy: 0.8021
Epoch 70: val_loss improved from 0.39657 to 0.39515, saving model to fashion_mnist_model.h5
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6071 - accuracy: 0.8021 - val_loss: 0.3951 - val_accuracy: 0.8681
Epoch 71/100
1685/1687 [============================>.] - ETA: 0s - loss: 0.5970 - accuracy: 0.8036
Epoch 71: val_loss did not improve from 0.39515
1687/1687 [==============================] - 13s 8ms/step - loss: 0.5972 - accuracy: 0.8035 - val_loss: 0.3957 - val_accuracy: 0.8703
Epoch 72/100
1685/1687 [============================>.] - ETA: 0s - loss: 0.6035 - accuracy: 0.8031
Epoch 72: val_loss did not improve from 0.39515
1687/1687 [==============================] - 14s 8ms/step - loss: 0.6037 - accuracy: 0.8030 - val_loss: 0.3984 - val_accuracy: 0.8658
Epoch 73/100
1686/1687 [============================>.] - ETA: 0s - loss: 0.6013 - accuracy: 0.8038
Epoch 73: val_loss did not improve from 0.39515
1687/1687 [==============================] - 13s 8ms/step - loss: 0.6013 - accuracy: 0.8038 - val_loss: 0.3977 - val_accuracy: 0.8651
Epoch 74/100
 683/1687 [===========>..................] - ETA: 6s - loss: 0.5873 - accuracy: 0.8068

# 코드 수정
acc = history.history['accuracy']
val_acc = history.history['val_accuracy']

loss = history.history['loss']
val_loss = history.history['val_loss']

epochs_range = range(len(acc))

plt.figure(figsize=(8, 8))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, acc, label='Training Accuracy')
plt.plot(epochs_range, val_acc, label='Validation Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Valid Accuracy')

plt.subplot(1, 2, 2)
plt.plot(epochs_range, loss, label='Training Loss')
plt.plot(epochs_range, val_loss, label='Validation Loss')
plt.legend(loc='upper right')
plt.title('Training and Valid Loss')
plt.show()

results = model.evaluate(test_dataset)