TensorFlow解説、Kerasチューナーの紹介
Kerasチューナーの紹介
https://www.tensorflow.org/tutorials/keras/keras_tuner
上記のTensorFlowチュートリアルを読んでつまづいたところのメモです。
img_train = img_train.astype('float32') / 255.0とは？
# Normalize pixel values between 0 and 1
img_train = img_train.astype('float32') / 255.0
img_test = img_test.astype('float32') / 255.0astype('float32')
は型変換をしてるんだろうけど、元の型は何だ？
img_train.dtypedtype('uint8')元の型はuint8でした。
x = img_train/255.0
x.dtypedtype('float64')単純に255.0で割り算するとfloat64になるみたいです。
# Normalize pixel values between 0 and 1
img_train = img_train.astype('float32') / 255.0
img_test = img_test.astype('float32') / 255.0
img_train.dtypedtype('float32')astype('float32')
を使うことでデータ型がfloat32になっています。
float64よりfloat32の方がメモリ消費が少なくて計算速度が速くて都合がいいのでわざわざ型変換しているんでしょう。
hpって何だ？
def model_builder(hp):
model = tf.keras.Sequential()
model.add(tf.keras.layers.Flatten(input_shape=(28, 28)))
# Tune the number of units in the first Dense layer
# Choose an optimal value between 32-512
hp_units = hp.Int('units', min_value=32, max_value=512, step=32)
model.add(tf.keras.layers.Dense(units=hp_units, activation='relu'))
model.add(tf.keras.layers.Dense(10))
# Tune the learning rate for the optimizer
# Choose an optimal value from 0.01, 0.001, or 0.0001
hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=hp_learning_rate),
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
return model
tuner = kt.Hyperband(model_builder,
objective='val_accuracy',
max_epochs=10,
factor=3,
directory='my_dir',
project_name='intro_to_kt')model_builder(hp)
の引数にあるhpって何だ？
たぶんHyper Parameterの略です。
tuner = kt.Hyperband(model_builder)
が呼ばれるときに
kt.Hyperband
から
model_builder
に何かが渡されるみたいです。
hp.Int('units', min_value=32, max_value=512, step=32)って何だ？
hp_units = hp.Int('units', min_value=32, max_value=512, step=32)
model.add(tf.keras.layers.Dense(units=hp_units, activation='relu'))映画レビューのテキスト分類
たとえば上記のTFチュートリアルでは
model.add(keras.layers.Dense(16, activation='relu'))
というのが登場しました。
これを省略せずに書くと
model.add(keras.layers.Dense(units=16, activation='relu'))
となります。
hp_units = hp.Int('units', min_value=32, max_value=512, step=32)
model.add(tf.keras.layers.Dense(units=hp_units, activation='relu'))
model.add(tf.keras.layers.Dense(units=32, activation='relu'))
でモデルを作成する。
model.add(tf.keras.layers.Dense(units=64, activation='relu'))
でモデルを作成する。
model.add(tf.keras.layers.Dense(units=96, activation='relu'))
でモデルを作成する。
という感じで、
model.add(tf.keras.layers.Dense(units=512, activation='relu'))
になるまで複数バージョンのモデルを作成してくれます。
色んなバージョンのモデルを生成して、どの設定のモデルが一番性能がいいかを後で選別します。
hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])って何だ？
hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=hp_learning_rate),
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])1e-2
というのは
1 × 10の-2乗という意味です。
つまり、
0.01
のことです。
1e-2 ＝ 0.01
1e-3 ＝ 0.001
1e-4 ＝ 0.0001
hp.Int()は最小値、最大値、ステップを渡すと設定を複数作成できましたが
hp.Choice()の場合は実際の選択肢をリストで渡します。
映画レビューのテキスト分類
上記のTFチュートリアルでは
model.compile(optimizer='Adam')
のような形式になっていた部分が今回は
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=hp_learning_rate))
のような形式になっています。
optimizer='Adam'
という指定の仕方だと学習率が設定できません。
optimizer=tf.keras.optimizers.Adam(learning_rate=hp_learning_rate)
という指定の仕方だと学習率が設定できます。
hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
を使って学習率を0.01, 0.001, 0.0001の3つのhp.Choiceにしているので
optimizer=tf.keras.optimizers.Adam(learning_rate=0.01)
という設定でモデルを作成、
optimizer=tf.keras.optimizers.Adam(learning_rate=0.001)
という設定でモデルを作成、
optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001)
という設定でモデルを作成、
という感じで複数のモデルが作成されます。
kt.Hyperband()とは？
def model_builder(hp):
model = tf.keras.Sequential()
model.add(tf.keras.layers.Flatten(input_shape=(28, 28)))
# Tune the number of units in the first Dense layer
# Choose an optimal value between 32-512
hp_units = hp.Int('units', min_value=32, max_value=512, step=32)
model.add(tf.keras.layers.Dense(units=hp_units, activation='relu'))
model.add(tf.keras.layers.Dense(10))
# Tune the learning rate for the optimizer
# Choose an optimal value from 0.01, 0.001, or 0.0001
hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=hp_learning_rate),
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
return model
tuner = kt.Hyperband(model_builder,
objective='val_accuracy',
max_epochs=10,
factor=3,
directory='my_dir',
project_name='intro_to_kt')kt.Hyperband()
を使うと複数のハイパーパラメーターを組み合わせて一番性能がいいモデルを自動的に探すことができます。
上記のコードだと、
Densuのユニット数が下記の15パターン
[32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, 480]
学習率が下記の3パターン
[0.01, 0,001, 0.0001]
ユニット数と学習率を組み合わせると45パターンもあります。
普通は
model.add(tf.keras.layers.Dense(units=32, activation='relu'))
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.01)
という組み合わせでモデルを作成して、
model.fit()
で学習して
accuracy
を算出。
model.add(tf.keras.layers.Dense(units=32, activation='relu'))
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.001)
という組み合わせでモデルを作成して、
model.fit()
で学習して
accuracy
を算出。
というのを45回繰り返してどの組み合わせが一番性能がいいかを調べます。
それを自動化するのが
kt.Hyperband()
の役目です。
tuner.search()の過程を見る方法
# tuner.search(img_train, label_train, epochs=10, validation_data=(img_test, label_test), callbacks=[ClearTrainingOutput()])
tuner.search(img_train, label_train, epochs=10, validation_data=(img_test, label_test))チュートリアルの中で出てくる
tuner.search(img_train, label_train, epochs=10, validation_data=(img_test, label_test), callbacks=[ClearTrainingOutput()])
から
callbacks=[ClearTrainingOutput()]
の部分を削除して
tuner.search(img_train, label_train, epochs=10, validation_data=(img_test, label_test))
としてやると、ハイパーパラメーター探索の過程を確認できます。
ユニット数と学習率の組み合わせ的には45通りあるはずなのに、実際には30通りしか探索されていませんでした。
Hyperbandの仕組み
Hyperbandのアルゴリズムが発明されるまでには大まかに3段階のステップがありました。
その１、原始的な方法。
ユニット数と学習率の組み合わせ的には45通りあります。
それぞれ10エポックずつ学習して、モデルの正解率が一番高いものを採用します。
このとき全部で450エポック分の学習が必要になります。
めっちゃ時間がかかる……
その２、Successive Halving
halveはhalfの動詞形です。
日本語でいうなら「連続半分法」
450エポックも学習すると時間がかかるから成績が悪いものは途中で枝刈します。
45通りのモデルで2エポックずつ学習。
正解率の高い22モデルを暫定採用、残りの23モデルは学習をストップ。
暫定採用の22モデルでさらに2エポックずつ学習。
正解率の高い11モデルを暫定採用、残りの11モデルは学習をストップ。
こんな感じでモデルを半分ずつに減らしていって、最終的に残った奴が最優秀モデルとして採用されます。
全部を10エポックずつ学習するよりも時間がかからないというメリットがあります。
その３、Hyperband
Successive Halvingの改良版。
Successive Halvingだと、エポックが進めば正解率が高くなるはずだったのに序盤で性能が出なかったために切り捨てられてしまう場合があります。
そこで、ランダムにモデルを選択しながら、切り捨てた分ももうちょっとだけエポックを進めて学習してみようというアルゴリズムです。
詳しくは元の論文を読みましょう。
Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization
kt.Hyperband()のパラメーター
tuner = kt.Hyperband(model_builder,
objective = 'val_accuracy',
max_epochs = 10,
factor = 3,
directory = 'my_dir',
project_name = 'intro_to_kt')directory = 'my_dir',
project_name = 'intro_to_kt'
の部分について、
my_dir\intro_to_kt
というディレクトリにハイパーパラメーター探索結果がファイルとして保存されます。
factorって何だ？
tuner = kt.Hyperband(model_builder,
objective = 'val_accuracy',
max_epochs = 10,
factor = 3,
directory = 'my_dir',
project_name = 'intro_to_kt') kt.Hyperband()
の引数にあるfactorって何だ？
ヘルプには
factor: Int. Reduction factor for the number of epochs
and number of models for each bracket.
とあります。
Tuners
https://keras-team.github.io/keras-tuner/documentation/tuners/#hyperband-class
kerastuner.tuners.hyperband.Hyperband(hypermodel, objective, max_epochs, factor=3, hyperband_iterations=1, seed=None, hyperparameters=None, tune_new_entries=True, allow_new_entries=True, **kwargs)
Hyperband
の引数
factor
の初期値は3とありますね。
Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization
factorは、元の論文に登場するηのことっぽい。
仮にハイパーパラメーターの組み合わせが72通りあった場合に
factor=2
なら、
72→36→18というように1/2ずつモデル数を絞り込んでいくのが
factor=3にすると
72→24→8というように1/3ずつモデルの数を絞り込んでいく。
だいたいこんな感じのイメージ。
overwrite=Trueを使う
tuner = kt.Hyperband(model_builder,
objective='val_accuracy',
max_epochs=10,
factor=3,
directory='my_dir',
project_name='intro_to_kt',
overwrite=True,
)kt.Hyperband()
を実行すると
my_dir/intro_to_kt
というディレクトリにいろいろとファイルが保存されます。
動作確認のために
kt.Hyperband()
を再度実行しようとすると
my_dir/intro_to_kt
に既に結果が保存されていて、そちらが読み込まれてしまいます。
探索済みのハイパーパラメーターを再利用するための機能ですが、探索の過程を把握するためにもう一度処理を実行したい。
そういうときは、
kt.Hyperband()
のオプションに
overwrite=True
を加えると、一から再実行できます。
Hyperbandの探索の過程を確認する（factor=2）
def model_builder(hp):
model = tf.keras.Sequential()
model.add(tf.keras.layers.Flatten(input_shape=(28, 28)))
# Tune the number of units in the first Dense layer
# Choose an optimal value between 32-512
hp_units = hp.Int('units', min_value=32, max_value=512, step=32)
model.add(tf.keras.layers.Dense(units=hp_units, activation='relu'))
model.add(tf.keras.layers.Dense(10))
# Tune the learning rate for the optimizer
# Choose an optimal value from 0.01, 0.001, or 0.0001
hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=hp_learning_rate),
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
return model
tuner = kt.Hyperband(model_builder,
objective='val_accuracy',
max_epochs=10,
factor=2,
directory='my_dir',
project_name='intro_to_kt',
overwrite=True,
)
tuner.search(img_train, label_train, epochs=10, validation_data=(img_test, label_test))kt.Hyperband()
に
factor=2
のオプションを指定。
tuner.search()
から
callbacks=[ClearTrainingOutput()]
を削除。
すると実行結果として以下のようなものが画面に表示されます。
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6656 - accuracy: 0.7630 - val_loss: 0.5888 - val_accuracy: 0.7873
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.4675 - accuracy: 0.8319 - val_loss: 0.4632 - val_accuracy: 0.8403
Trial complete
Trial summary
|-Trial ID: 9cdc1c442f71ba866e16b060f0bdc3b2
|-Score: 0.8403000235557556
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 448
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6843 - accuracy: 0.7584 - val_loss: 0.5602 - val_accuracy: 0.7941
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.5145 - accuracy: 0.8145 - val_loss: 0.6000 - val_accuracy: 0.7929
Trial complete
Trial summary
|-Trial ID: 81d1ead5d4755eb784c030c4f65ff80e
|-Score: 0.7940999865531921
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 64
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7010 - accuracy: 0.7507 - val_loss: 0.5235 - val_accuracy: 0.8131
Epoch 2/2
313/313 [==============================] - 1s 3ms/step - loss: 0.5019 - accuracy: 0.8191 - val_loss: 0.5466 - val_accuracy: 0.8034
Trial complete
Trial summary
|-Trial ID: cbaae64ea0b187c8f38e934f71e88491
|-Score: 0.8130999803543091
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 96
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7520 - accuracy: 0.7497 - val_loss: 0.7055 - val_accuracy: 0.7556
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.4947 - accuracy: 0.8229 - val_loss: 0.5171 - val_accuracy: 0.8171
Trial complete
Trial summary
|-Trial ID: 2abca7a8074f185137aeee09f95b63ed
|-Score: 0.8170999884605408
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 224
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6936 - accuracy: 0.7562 - val_loss: 0.5733 - val_accuracy: 0.7959
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.4784 - accuracy: 0.8290 - val_loss: 0.5017 - val_accuracy: 0.8198
Trial complete
Trial summary
|-Trial ID: 561925038c121c34b35fc2886bd7db5d
|-Score: 0.8198000192642212
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 224
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.0544 - accuracy: 0.6656 - val_loss: 0.7515 - val_accuracy: 0.7599
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6498 - accuracy: 0.7898 - val_loss: 0.6252 - val_accuracy: 0.7897
Trial complete
Trial summary
|-Trial ID: 2fb674f99ff4085b0f83236c6d9501a0
|-Score: 0.7896999716758728
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 288
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.0207 - accuracy: 0.6802 - val_loss: 0.7121 - val_accuracy: 0.7673
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6242 - accuracy: 0.7973 - val_loss: 0.5991 - val_accuracy: 0.7988
Trial complete
Trial summary
|-Trial ID: 10bda8ed6b592a44ba79c59ac07b5140
|-Score: 0.798799991607666
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 384
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7711 - accuracy: 0.7459 - val_loss: 0.6375 - val_accuracy: 0.7709
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.5168 - accuracy: 0.8112 - val_loss: 0.5474 - val_accuracy: 0.8124
Trial complete
Trial summary
|-Trial ID: 1ad2728d3233e0f636dbf04210d12405
|-Score: 0.8123999834060669
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 480
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.9001 - accuracy: 0.6930 - val_loss: 0.6270 - val_accuracy: 0.7759
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.5550 - accuracy: 0.8072 - val_loss: 0.5446 - val_accuracy: 0.8098
Trial complete
Trial summary
|-Trial ID: 86090a6e7308b9495b3e4550135c8abc
|-Score: 0.8098000288009644
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 32
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6650 - accuracy: 0.7646 - val_loss: 0.5602 - val_accuracy: 0.8063
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.4720 - accuracy: 0.8341 - val_loss: 0.5134 - val_accuracy: 0.8195
Trial complete
Trial summary
|-Trial ID: c117e736441cfa852a4cd35dc9e0fde4
|-Score: 0.8195000290870667
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 384
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6615 - accuracy: 0.7671 - val_loss: 0.5450 - val_accuracy: 0.8102
Trial complete
Trial summary
|-Trial ID: 7f64f5afd9946e0098c4efaec54dce90
|-Score: 0.8101999759674072
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 9cdc1c442f71ba866e16b060f0bdc3b2
|-units: 448
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6829 - accuracy: 0.7593 - val_loss: 0.5728 - val_accuracy: 0.7882
Trial complete
Trial summary
|-Trial ID: b1acfcbfb5e10d50132e7c32cc1726fe
|-Score: 0.7882000207901001
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 561925038c121c34b35fc2886bd7db5d
|-units: 224
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6825 - accuracy: 0.7615 - val_loss: 0.5352 - val_accuracy: 0.8039
Trial complete
Trial summary
|-Trial ID: a21714cffbdcc375e6165f85ff3825c0
|-Score: 0.8039000034332275
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: c117e736441cfa852a4cd35dc9e0fde4
|-units: 384
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7334 - accuracy: 0.7540 - val_loss: 0.5880 - val_accuracy: 0.7748
Trial complete
Trial summary
|-Trial ID: 76e88295170ec73ef9c38da2a2d9b1b8
|-Score: 0.7748000025749207
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 2abca7a8074f185137aeee09f95b63ed
|-units: 224
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7065 - accuracy: 0.7592 - val_loss: 0.6044 - val_accuracy: 0.7909
Trial complete
Trial summary
|-Trial ID: d5c7aff2a720c4b5f6523cccc38e79a7
|-Score: 0.7908999919891357
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: cbaae64ea0b187c8f38e934f71e88491
|-units: 96
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6648 - accuracy: 0.7681 - val_loss: 0.5311 - val_accuracy: 0.8177
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4750 - accuracy: 0.8329 - val_loss: 0.4802 - val_accuracy: 0.8283
Trial complete
Trial summary
|-Trial ID: 7b4a1ddad48a84d59de8670aa83fb3c1
|-Score: 0.8282999992370605
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 2
|-tuner/trial_id: 7f64f5afd9946e0098c4efaec54dce90
|-units: 448
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6666 - accuracy: 0.7662 - val_loss: 0.5422 - val_accuracy: 0.8113
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4692 - accuracy: 0.8352 - val_loss: 0.5543 - val_accuracy: 0.7900
Trial complete
Trial summary
|-Trial ID: 0e41daef7808040736c9fed760a84e84
|-Score: 0.8112999796867371
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 2
|-tuner/trial_id: a21714cffbdcc375e6165f85ff3825c0
|-units: 384
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6973 - accuracy: 0.7489 - val_loss: 0.5374 - val_accuracy: 0.8080
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4960 - accuracy: 0.8205 - val_loss: 0.4936 - val_accuracy: 0.8263
Trial complete
Trial summary
|-Trial ID: 51859df3c1fc864381a46544adcd6418
|-Score: 0.8263000249862671
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 2
|-tuner/trial_id: d5c7aff2a720c4b5f6523cccc38e79a7
|-units: 96
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6569 - accuracy: 0.7691 - val_loss: 0.5598 - val_accuracy: 0.8064
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4702 - accuracy: 0.8344 - val_loss: 0.5364 - val_accuracy: 0.8139
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4107 - accuracy: 0.8509 - val_loss: 0.4511 - val_accuracy: 0.8401
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3690 - accuracy: 0.8653 - val_loss: 0.5122 - val_accuracy: 0.8147
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3471 - accuracy: 0.8710 - val_loss: 0.4645 - val_accuracy: 0.8240
Trial complete
Trial summary
|-Trial ID: e25c5575878457f71502e274e9cb2547
|-Score: 0.8400999903678894
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 3
|-tuner/trial_id: 7b4a1ddad48a84d59de8670aa83fb3c1
|-units: 448
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7192 - accuracy: 0.7494 - val_loss: 0.6321 - val_accuracy: 0.7783
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5126 - accuracy: 0.8125 - val_loss: 0.5175 - val_accuracy: 0.8182
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4652 - accuracy: 0.8275 - val_loss: 0.5493 - val_accuracy: 0.8082
Epoch 9/10
313/313 [==============================] - 1s 3ms/step - loss: 0.4282 - accuracy: 0.8419 - val_loss: 0.5467 - val_accuracy: 0.8060
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4072 - accuracy: 0.8494 - val_loss: 0.4686 - val_accuracy: 0.8433
Trial complete
Trial summary
|-Trial ID: ec7818f7e256c3654331d8b1dd0a7458
|-Score: 0.8432999849319458
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 3
|-tuner/trial_id: 51859df3c1fc864381a46544adcd6418
|-units: 96
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 1.0801 - accuracy: 0.6611 - val_loss: 0.7417 - val_accuracy: 0.7575
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6602 - accuracy: 0.7840 - val_loss: 0.6360 - val_accuracy: 0.7848
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5717 - accuracy: 0.8095 - val_loss: 0.5815 - val_accuracy: 0.8037
Trial complete
Trial summary
|-Trial ID: 72377b0c317f41736054cde3676a271a
|-Score: 0.8036999702453613
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 256
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 1.1268 - accuracy: 0.6553 - val_loss: 0.7977 - val_accuracy: 0.7362
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6924 - accuracy: 0.7801 - val_loss: 0.6646 - val_accuracy: 0.7789
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5991 - accuracy: 0.8058 - val_loss: 0.6066 - val_accuracy: 0.7976
Trial complete
Trial summary
|-Trial ID: 00d689a890bd957605a9ca5fec33e632
|-Score: 0.7975999712944031
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 160
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7641 - accuracy: 0.7341 - val_loss: 0.5656 - val_accuracy: 0.8052
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5099 - accuracy: 0.8224 - val_loss: 0.5141 - val_accuracy: 0.8212
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4670 - accuracy: 0.8330 - val_loss: 0.4769 - val_accuracy: 0.8315
Trial complete
Trial summary
|-Trial ID: b3d2b6a8f435ec2648714d2d6db89b20
|-Score: 0.8314999938011169
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 64
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6960 - accuracy: 0.7564 - val_loss: 0.5658 - val_accuracy: 0.7938
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5088 - accuracy: 0.8165 - val_loss: 0.5533 - val_accuracy: 0.8071
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4488 - accuracy: 0.8346 - val_loss: 0.5292 - val_accuracy: 0.8137
Trial complete
Trial summary
|-Trial ID: 323c2efbe7d4acd3e14b3945871c5ad9
|-Score: 0.8137000203132629
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 2
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 160
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6758 - accuracy: 0.7646 - val_loss: 0.5350 - val_accuracy: 0.8182
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4857 - accuracy: 0.8311 - val_loss: 0.5108 - val_accuracy: 0.8176
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4197 - accuracy: 0.8539 - val_loss: 0.4540 - val_accuracy: 0.8390
Trial complete
Trial summary
|-Trial ID: 3f731b89b7114269fc4a9f99b8b28628
|-Score: 0.8389999866485596
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 192
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.9729 - accuracy: 0.6970 - val_loss: 0.6887 - val_accuracy: 0.7695
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6004 - accuracy: 0.8000 - val_loss: 0.5857 - val_accuracy: 0.8048
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5261 - accuracy: 0.8232 - val_loss: 0.5394 - val_accuracy: 0.8151
Trial complete
Trial summary
|-Trial ID: 0f57bdce076be831e4ee0012919eed28
|-Score: 0.8151000142097473
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 512
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 1.0164 - accuracy: 0.6853 - val_loss: 0.7173 - val_accuracy: 0.7560
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6262 - accuracy: 0.7944 - val_loss: 0.6018 - val_accuracy: 0.7968
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5455 - accuracy: 0.8189 - val_loss: 0.5703 - val_accuracy: 0.7992
Trial complete
Trial summary
|-Trial ID: 52aecb39e48d9f4b9107a7e5c84200a4
|-Score: 0.7991999983787537
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 352
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6943 - accuracy: 0.7578 - val_loss: 0.5949 - val_accuracy: 0.7821
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4801 - accuracy: 0.8298 - val_loss: 0.5124 - val_accuracy: 0.8119
Trial complete
Trial summary
|-Trial ID: aa757c6f27aad793873f7298602c48fe
|-Score: 0.8119000196456909
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: 3f731b89b7114269fc4a9f99b8b28628
|-units: 192
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7597 - accuracy: 0.7404 - val_loss: 0.5675 - val_accuracy: 0.7998
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5059 - accuracy: 0.8196 - val_loss: 0.5133 - val_accuracy: 0.8170
Trial complete
Trial summary
|-Trial ID: 090f4bc9f99f647c8fbd24236c570dca
|-Score: 0.8169999718666077
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: b3d2b6a8f435ec2648714d2d6db89b20
|-units: 64
Epoch 4/5
1/313 [..............................] - ETA: 0s - loss: 2.3192 - accuracy: 0.1562WARNING:tensorflow:Callbacks method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0019s vs `on_train_batch_end` time: 0.0032s). Check your callbacks.
313/313 [==============================] - 1s 4ms/step - loss: 0.9532 - accuracy: 0.6997 - val_loss: 0.6733 - val_accuracy: 0.7684
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5973 - accuracy: 0.8061 - val_loss: 0.5800 - val_accuracy: 0.8080
Trial complete
Trial summary
|-Trial ID: 24600d23b69dda5319278bef4ec7129d
|-Score: 0.8080000281333923
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: 0f57bdce076be831e4ee0012919eed28
|-units: 512
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7172 - accuracy: 0.7497 - val_loss: 0.5907 - val_accuracy: 0.7840
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5237 - accuracy: 0.8074 - val_loss: 0.5161 - val_accuracy: 0.8130
Trial complete
Trial summary
|-Trial ID: 6349877acb94dbcdb2c3902b40f79fe1
|-Score: 0.8130000233650208
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: 323c2efbe7d4acd3e14b3945871c5ad9
|-units: 160
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7924 - accuracy: 0.7254 - val_loss: 0.6004 - val_accuracy: 0.7834
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5149 - accuracy: 0.8236 - val_loss: 0.5047 - val_accuracy: 0.8266
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4528 - accuracy: 0.8418 - val_loss: 0.4758 - val_accuracy: 0.8354
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4172 - accuracy: 0.8542 - val_loss: 0.4881 - val_accuracy: 0.8257
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3939 - accuracy: 0.8639 - val_loss: 0.5125 - val_accuracy: 0.8174
Trial complete
Trial summary
|-Trial ID: 40186a93f2086bf6249d1e571c746789
|-Score: 0.8353999853134155
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 2
|-tuner/trial_id: 090f4bc9f99f647c8fbd24236c570dca
|-units: 64
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7434 - accuracy: 0.7342 - val_loss: 0.5904 - val_accuracy: 0.7796
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5084 - accuracy: 0.8153 - val_loss: 0.5270 - val_accuracy: 0.8032
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4475 - accuracy: 0.8335 - val_loss: 0.5769 - val_accuracy: 0.7890
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4152 - accuracy: 0.8453 - val_loss: 0.4931 - val_accuracy: 0.8242
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4089 - accuracy: 0.8500 - val_loss: 0.5188 - val_accuracy: 0.8239
Trial complete
Trial summary
|-Trial ID: 91693f72e3121824f1049c2e30d9c59b
|-Score: 0.8241999745368958
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 2
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 2
|-tuner/trial_id: 6349877acb94dbcdb2c3902b40f79fe1
|-units: 160
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7494 - accuracy: 0.7374 - val_loss: 0.5799 - val_accuracy: 0.7976
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5051 - accuracy: 0.8156 - val_loss: 0.5305 - val_accuracy: 0.8132
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4646 - accuracy: 0.8321 - val_loss: 0.5019 - val_accuracy: 0.8193
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4305 - accuracy: 0.8409 - val_loss: 0.5494 - val_accuracy: 0.8053
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4170 - accuracy: 0.8479 - val_loss: 0.5292 - val_accuracy: 0.8247
Trial complete
Trial summary
|-Trial ID: 09ff3a700e2d58e102bc49936a0de5b3
|-Score: 0.8246999979019165
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 1
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 288
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6904 - accuracy: 0.7560 - val_loss: 0.5360 - val_accuracy: 0.8074
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5193 - accuracy: 0.8122 - val_loss: 0.5048 - val_accuracy: 0.8229
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4692 - accuracy: 0.8279 - val_loss: 0.5969 - val_accuracy: 0.7951
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4234 - accuracy: 0.8483 - val_loss: 0.5021 - val_accuracy: 0.8210
Epoch 5/5
313/313 [==============================] - 1s 3ms/step - loss: 0.4021 - accuracy: 0.8532 - val_loss: 0.5773 - val_accuracy: 0.8007
Trial complete
Trial summary
|-Trial ID: 6ccd071cf55dfdc622cca921ab0eadc8
|-Score: 0.8228999972343445
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 1
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 32
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7551 - accuracy: 0.7511 - val_loss: 0.5894 - val_accuracy: 0.7867
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4933 - accuracy: 0.8188 - val_loss: 0.5302 - val_accuracy: 0.8024
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4379 - accuracy: 0.8369 - val_loss: 0.5591 - val_accuracy: 0.7955
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4210 - accuracy: 0.8425 - val_loss: 0.5439 - val_accuracy: 0.8105
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4025 - accuracy: 0.8568 - val_loss: 0.4962 - val_accuracy: 0.8334
Trial complete
Trial summary
|-Trial ID: e7ee048827ac8c80fc957e0ba2ea297c
|-Score: 0.8334000110626221
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 1
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 256
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7560 - accuracy: 0.7494 - val_loss: 0.5592 - val_accuracy: 0.8019
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4938 - accuracy: 0.8234 - val_loss: 0.5297 - val_accuracy: 0.8218
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4452 - accuracy: 0.8397 - val_loss: 0.5171 - val_accuracy: 0.8192
Epoch 4/5
313/313 [==============================] - 1s 3ms/step - loss: 0.4393 - accuracy: 0.8400 - val_loss: 0.5470 - val_accuracy: 0.8162
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4283 - accuracy: 0.8435 - val_loss: 0.5150 - val_accuracy: 0.8296
Trial complete
Trial summary
|-Trial ID: ca023520db336d1b0c0bff3a90640b60
|-Score: 0.8295999765396118
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 1
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 384
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6909 - accuracy: 0.7585 - val_loss: 0.6049 - val_accuracy: 0.7797
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4847 - accuracy: 0.8281 - val_loss: 0.5030 - val_accuracy: 0.8227
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4348 - accuracy: 0.8480 - val_loss: 0.4669 - val_accuracy: 0.8367
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.3890 - accuracy: 0.8599 - val_loss: 0.4390 - val_accuracy: 0.8445
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.3658 - accuracy: 0.8690 - val_loss: 0.4378 - val_accuracy: 0.8464
Trial complete
Trial summary
|-Trial ID: 63e4f4fcaed7552395e08794ab70afed
|-Score: 0.8464000225067139
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 1
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 160
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7054 - accuracy: 0.7523 - val_loss: 0.6246 - val_accuracy: 0.7750
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4815 - accuracy: 0.8333 - val_loss: 0.5283 - val_accuracy: 0.8056
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4222 - accuracy: 0.8520 - val_loss: 0.4704 - val_accuracy: 0.8354
Epoch 9/10
313/313 [==============================] - 1s 3ms/step - loss: 0.3988 - accuracy: 0.8591 - val_loss: 0.4604 - val_accuracy: 0.8347
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3570 - accuracy: 0.8714 - val_loss: 0.4324 - val_accuracy: 0.8449
Trial complete
Trial summary
|-Trial ID: 4fba0e0f2466ddcd1a60d4ac0d49b373
|-Score: 0.8449000120162964
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 1
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 1
|-tuner/trial_id: 63e4f4fcaed7552395e08794ab70afed
|-units: 160
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7365 - accuracy: 0.7466 - val_loss: 0.6144 - val_accuracy: 0.7859
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5271 - accuracy: 0.8094 - val_loss: 0.5079 - val_accuracy: 0.8167
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4795 - accuracy: 0.8283 - val_loss: 0.5564 - val_accuracy: 0.7987
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4430 - accuracy: 0.8397 - val_loss: 0.5467 - val_accuracy: 0.8139
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4029 - accuracy: 0.8566 - val_loss: 0.5117 - val_accuracy: 0.8340
Trial complete
Trial summary
|-Trial ID: 6d2da1cc2c5035efa50c39dbb62c6b3f
|-Score: 0.8339999914169312
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 1
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 1
|-tuner/trial_id: e7ee048827ac8c80fc957e0ba2ea297c
|-units: 256
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7891 - accuracy: 0.7433 - val_loss: 0.5541 - val_accuracy: 0.8016
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5067 - accuracy: 0.8155 - val_loss: 0.5630 - val_accuracy: 0.7976
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4706 - accuracy: 0.8301 - val_loss: 0.5281 - val_accuracy: 0.8140
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4327 - accuracy: 0.8402 - val_loss: 0.5770 - val_accuracy: 0.8042
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4090 - accuracy: 0.8478 - val_loss: 0.5547 - val_accuracy: 0.8104
Trial complete
Trial summary
|-Trial ID: 83e038269ab8d14736b2594fa94968ac
|-Score: 0.8140000104904175
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 1
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 1
|-tuner/trial_id: ca023520db336d1b0c0bff3a90640b60
|-units: 384
Epoch 1/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6575 - accuracy: 0.7671 - val_loss: 0.5340 - val_accuracy: 0.8074
Epoch 2/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4765 - accuracy: 0.8283 - val_loss: 0.5010 - val_accuracy: 0.8233
Epoch 3/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4040 - accuracy: 0.8526 - val_loss: 0.4814 - val_accuracy: 0.8290
Epoch 4/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3749 - accuracy: 0.8626 - val_loss: 0.4515 - val_accuracy: 0.8442
Epoch 5/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3358 - accuracy: 0.8774 - val_loss: 0.4417 - val_accuracy: 0.8443
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3084 - accuracy: 0.8877 - val_loss: 0.4164 - val_accuracy: 0.8588
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.2928 - accuracy: 0.8948 - val_loss: 0.4261 - val_accuracy: 0.8539
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.2841 - accuracy: 0.8956 - val_loss: 0.4513 - val_accuracy: 0.8474
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.2675 - accuracy: 0.8997 - val_loss: 0.4494 - val_accuracy: 0.8449
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.2613 - accuracy: 0.9035 - val_loss: 0.4471 - val_accuracy: 0.8512
Trial complete
Trial summary
|-Trial ID: e158aaae79a9d16ce132ba20261debf5
|-Score: 0.8587999939918518
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 0
|-tuner/epochs: 10
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 480
Epoch 1/10
313/313 [==============================] - 1s 4ms/step - loss: 1.4031 - accuracy: 0.5836 - val_loss: 0.9521 - val_accuracy: 0.7018
Epoch 2/10
313/313 [==============================] - 1s 4ms/step - loss: 0.8212 - accuracy: 0.7390 - val_loss: 0.7571 - val_accuracy: 0.7430
Epoch 3/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6935 - accuracy: 0.7704 - val_loss: 0.6758 - val_accuracy: 0.7666
Epoch 4/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6278 - accuracy: 0.7910 - val_loss: 0.6297 - val_accuracy: 0.7846
Epoch 5/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5842 - accuracy: 0.8066 - val_loss: 0.5988 - val_accuracy: 0.7956
Epoch 6/10
313/313 [==============================] - 1s 3ms/step - loss: 0.5535 - accuracy: 0.8150 - val_loss: 0.5712 - val_accuracy: 0.8032
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5274 - accuracy: 0.8224 - val_loss: 0.5563 - val_accuracy: 0.8098
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5067 - accuracy: 0.8301 - val_loss: 0.5423 - val_accuracy: 0.8152
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4902 - accuracy: 0.8350 - val_loss: 0.5275 - val_accuracy: 0.8211
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4759 - accuracy: 0.8391 - val_loss: 0.5195 - val_accuracy: 0.8218
Trial complete
Trial summary
|-Trial ID: 603210b96528b49e1246c234fc2dcffa
|-Score: 0.8217999935150146
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 0
|-tuner/epochs: 10
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 64
Epoch 1/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7118 - accuracy: 0.7495 - val_loss: 0.6189 - val_accuracy: 0.7762
Epoch 2/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4917 - accuracy: 0.8309 - val_loss: 0.5180 - val_accuracy: 0.8175
Epoch 3/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4409 - accuracy: 0.8469 - val_loss: 0.5116 - val_accuracy: 0.8188
Epoch 4/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3968 - accuracy: 0.8594 - val_loss: 0.5112 - val_accuracy: 0.8161
Epoch 5/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3673 - accuracy: 0.8702 - val_loss: 0.4573 - val_accuracy: 0.8344
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3542 - accuracy: 0.8718 - val_loss: 0.4384 - val_accuracy: 0.8466
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3298 - accuracy: 0.8834 - val_loss: 0.4220 - val_accuracy: 0.8542
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3018 - accuracy: 0.8917 - val_loss: 0.4419 - val_accuracy: 0.8445
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.2852 - accuracy: 0.8993 - val_loss: 0.4323 - val_accuracy: 0.8504
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.2832 - accuracy: 0.8965 - val_loss: 0.4190 - val_accuracy: 0.8538
Trial complete
Trial summary
|-Trial ID: d41de33e3b916d3dc1dd281a734242b9
|-Score: 0.854200005531311
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 0
|-tuner/epochs: 10
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 128
Epoch 1/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7935 - accuracy: 0.7455 - val_loss: 0.5937 - val_accuracy: 0.7725
Epoch 2/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4994 - accuracy: 0.8181 - val_loss: 0.5443 - val_accuracy: 0.8040
Epoch 3/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4637 - accuracy: 0.8303 - val_loss: 0.5416 - val_accuracy: 0.8041
Epoch 4/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4008 - accuracy: 0.8546 - val_loss: 0.4923 - val_accuracy: 0.8367
Epoch 5/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3812 - accuracy: 0.8614 - val_loss: 0.4872 - val_accuracy: 0.8341
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3795 - accuracy: 0.8617 - val_loss: 0.4754 - val_accuracy: 0.8453
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3696 - accuracy: 0.8650 - val_loss: 0.5289 - val_accuracy: 0.8221
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3414 - accuracy: 0.8733 - val_loss: 0.5208 - val_accuracy: 0.8386
Epoch 9/10
313/313 [==============================] - 1s 3ms/step - loss: 0.3398 - accuracy: 0.8721 - val_loss: 0.6307 - val_accuracy: 0.8215
Epoch 10/10
313/313 [==============================] - 1s 3ms/step - loss: 0.3279 - accuracy: 0.8818 - val_loss: 0.5430 - val_accuracy: 0.8210
Trial complete
Trial summary
|-Trial ID: 9f2a0c81807835bdee27530498841ae8
|-Score: 0.845300018787384
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 0
|-tuner/epochs: 10
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 416
Epoch 1/10
313/313 [==============================] - 1s 4ms/step - loss: 1.0264 - accuracy: 0.6748 - val_loss: 0.7149 - val_accuracy: 0.7672
Epoch 2/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6196 - accuracy: 0.7990 - val_loss: 0.6010 - val_accuracy: 0.7968
Epoch 3/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5400 - accuracy: 0.8189 - val_loss: 0.5461 - val_accuracy: 0.8167
Epoch 4/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4945 - accuracy: 0.8364 - val_loss: 0.5209 - val_accuracy: 0.8231
Epoch 5/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4622 - accuracy: 0.8431 - val_loss: 0.5014 - val_accuracy: 0.8304
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4452 - accuracy: 0.8462 - val_loss: 0.4900 - val_accuracy: 0.8333
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4192 - accuracy: 0.8610 - val_loss: 0.4827 - val_accuracy: 0.8306
Epoch 8/10
313/313 [==============================] - 1s 3ms/step - loss: 0.4054 - accuracy: 0.8626 - val_loss: 0.4818 - val_accuracy: 0.8316
Epoch 9/10
313/313 [==============================] - 1s 3ms/step - loss: 0.3941 - accuracy: 0.8664 - val_loss: 0.4920 - val_accuracy: 0.8267
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3811 - accuracy: 0.8702 - val_loss: 0.4633 - val_accuracy: 0.8387
Trial complete
Trial summary
|-Trial ID: f873d76442942683362abe9b5d8b32f0
|-Score: 0.838699996471405
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 0
|-tuner/epochs: 10
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 416
INFO:tensorflow:Oracle triggered exitbracket3
|-tuner/bracket: 3
|-tuner/epochs: 2
|-tuner/round: 0
ブラケット3、ラウンド0のときエポック数は2
この条件で10個のモデルが実行されています。
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/round: 1
ブラケット3、ラウンド1のときエポック数は3
この条件で5個のモデルが実行されています。
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/round: 2
ブラケット3、ラウンド2のときエポック数は5
この条件で3個のモデルが実行されています。
|-tuner/bracket: 3
|-tuner/epochs: 10
|-tuner/round: 3
ブラケット3、ラウンド3のときエポック数は10
この条件で2個のモデルが実行されています。
bracket2
|-tuner/bracket: 2
|-tuner/epochs: 3
|-tuner/round: 0
ブラケット2、ラウンド0のときエポック数は3
この条件で7個のモデルが実行されています。
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/round: 1
ブラケット2、ラウンド1のときエポック数は5
この条件で4個のモデルが実行されています。
|-tuner/bracket: 2
|-tuner/epochs: 10
|-tuner/round: 2
ブラケット2、ラウンド2のときエポック数は10
この条件で2個のモデルが実行されています。
bracket1
|-tuner/bracket: 1
|-tuner/epochs: 5
|-tuner/round: 0
ブラケット1、ラウンド0のときエポック数は5
この条件で5個のモデルが実行されています。
|-tuner/bracket: 1
|-tuner/epochs: 10
|-tuner/round: 1
ブラケット1、ラウンド1のときエポック数は10
この条件で3個のモデルが実行されています。
bracket0
|-tuner/bracket: 0
|-tuner/epochs: 10
|-tuner/round: 0
ブラケット0、ラウンド0のときエポック数は10
この条件で5個のモデルが実行されています。
Hyperbandの探索過程を整理
要約しましょう。
bracketというのが4→3→2→1→0という風に進んでいく。
各braketの中では
roundが0→1→2→……
と進んでいく。
これが全体の流れです。
braket3のとき、ラウンドが進むにつれて
エポック数が2→3→5→10と増える。
モデル数は10→5→3→2と減る。
今回のチュートリアルでは、ハイパーパラメータの組み合わせ的に45個のモデルがありました。
この中から10個だけがランダムに選ばれます。
その10個のモデルを2エポックまで学習して、正解率が高い5個に絞り込む。
5個のモデルを3エポックまで学習して、正解率が高い3個に絞り込む。
3個のモデルを5エポックまで学習して、正解率が高い2個に絞り込む。
2個のモデルを10エポックまで学習して、一番正解率が高いモデルを決定。
|-learning_rate: 0.01
|-units: 96
の組み合わせが
|-Score: 0.8432999849319458
で一番成績が良い
という結果が得られました。
これがbraket3でのできごと。
braket3
エポック数： 2→3→5→10
モデル数： 10→5→3→2
（45個あるモデルからランダムに10個選んで1つにまで絞り込んでいる）
bracket2
エポック数： 3→5→10
モデル数： 7→4→2
（45個あるモデルからランダムに7個選んで1つにまで絞り込んでいる）
bracket1
エポック数： 5→10
モデル数： 5→3
（45個あるモデルからランダムに5個選んで1つにまで絞り込んでいる）
bracket0
エポック数： 10
モデル数： 5
（45個あるモデルからランダムに5個選んで1つにまで絞り込んでいる）
braket3～braket0までの結果を総合して一番いいモデルを決定。
Hyperbandの内部では、こういう作業が行われています。
kt.Hyperband()
のオプションに
factor=2
を指定したことを思い出しましょう。
モデル数が
10→5→3→2
のように絞り込まれているのはfactor=2なので、1/2ずつ個数を減らしています。
factor=3だったら1/3ずつ絞り込まれていきます。
各bracketで最初のモデルはランダムに選ばれてるっぽいんだけど、45個あるモデルから10個しか選ばれないとかなので、1度も実行されないモデルが存在しますね。
bracket3で10個、
bracket2で7個、
bracket1で5個、
bracket0で5個、
各ブラケットの初期に選出されるモデルが被らなかったとしても27個のモデルしか探索されていない。
これだと最良のハイパーパラメーターを見つけられない可能性があるのでは？
factor=2、epochs=20で再検証
def model_builder(hp):
model = tf.keras.Sequential()
model.add(tf.keras.layers.Flatten(input_shape=(28, 28)))
# Tune the number of units in the first Dense layer
# Choose an optimal value between 32-512
hp_units = hp.Int('units', min_value=32, max_value=512, step=32)
model.add(tf.keras.layers.Dense(units=hp_units, activation='relu'))
model.add(tf.keras.layers.Dense(10))
# Tune the learning rate for the optimizer
# Choose an optimal value from 0.01, 0.001, or 0.0001
hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=hp_learning_rate),
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
return model
tuner = kt.Hyperband(model_builder,
objective='val_accuracy',
max_epochs=20,
factor=2,
directory='my_dir',
project_name='intro_to_kt',
overwrite=True,
)
tuner.search(img_train, label_train, epochs=20, validation_data=(img_test, label_test))kt.Hyperband()
の
max_epochs=10
のところを
max_epochs=20
に変更。
tuner.search()
の
epochs=10
のところを
epochs=20
に変更。
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6802 - accuracy: 0.7615 - val_loss: 0.5469 - val_accuracy: 0.8079
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.4767 - accuracy: 0.8315 - val_loss: 0.4925 - val_accuracy: 0.8271
Trial complete
Trial summary
|-Trial ID: 47f427b9ff04f357806160ab4ee25b81
|-Score: 0.8270999789237976
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 192
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7836 - accuracy: 0.7490 - val_loss: 0.5555 - val_accuracy: 0.8172
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.5041 - accuracy: 0.8176 - val_loss: 0.5410 - val_accuracy: 0.8212
Trial complete
Trial summary
|-Trial ID: 302193218a5e1690cfe753f1603f1737
|-Score: 0.8212000131607056
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 320
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.0339 - accuracy: 0.6857 - val_loss: 0.7293 - val_accuracy: 0.7676
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6415 - accuracy: 0.7944 - val_loss: 0.6214 - val_accuracy: 0.7915
Trial complete
Trial summary
|-Trial ID: 3a43f2df38c41bc8759791b647948110
|-Score: 0.7914999723434448
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 288
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6823 - accuracy: 0.7560 - val_loss: 0.5463 - val_accuracy: 0.8087
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.4682 - accuracy: 0.8346 - val_loss: 0.4695 - val_accuracy: 0.8356
Trial complete
Trial summary
|-Trial ID: 6180fcc92dc041b82ec3e4f2ade58b12
|-Score: 0.8356000185012817
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 384
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.3039 - accuracy: 0.5934 - val_loss: 0.8728 - val_accuracy: 0.7169
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7564 - accuracy: 0.7541 - val_loss: 0.7026 - val_accuracy: 0.7645
Trial complete
Trial summary
|-Trial ID: efe4fa7e5dd5cb496a8fd5a0c200190c
|-Score: 0.7645000219345093
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 96
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7184 - accuracy: 0.7399 - val_loss: 0.6963 - val_accuracy: 0.7475
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.5158 - accuracy: 0.8140 - val_loss: 0.5197 - val_accuracy: 0.8160
Trial complete
Trial summary
|-Trial ID: cd1ced6111b608f8bf8dd92d1362ade9
|-Score: 0.8159999847412109
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 64
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.9979 - accuracy: 0.6868 - val_loss: 0.7149 - val_accuracy: 0.7700
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6220 - accuracy: 0.7983 - val_loss: 0.5994 - val_accuracy: 0.7981
Trial complete
Trial summary
|-Trial ID: cb74ea960bfd9a5f81c5d96b23b8beb2
|-Score: 0.7980999946594238
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 352
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.3880 - accuracy: 0.5797 - val_loss: 0.9766 - val_accuracy: 0.6913
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.8344 - accuracy: 0.7338 - val_loss: 0.7717 - val_accuracy: 0.7465
Trial complete
Trial summary
|-Trial ID: b40dd4de415ba7999d09bdf9e821745b
|-Score: 0.7465000152587891
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 64
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6751 - accuracy: 0.7624 - val_loss: 0.5065 - val_accuracy: 0.8227
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.4739 - accuracy: 0.8311 - val_loss: 0.4843 - val_accuracy: 0.8224
Trial complete
Trial summary
|-Trial ID: 4bc37ea30b559871e97cb13fb9884507
|-Score: 0.822700023651123
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 320
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7949 - accuracy: 0.7407 - val_loss: 0.7076 - val_accuracy: 0.7502
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.5170 - accuracy: 0.8138 - val_loss: 0.5263 - val_accuracy: 0.8133
Trial complete
Trial summary
|-Trial ID: c24688f868fcd6c5056e05846148bb68
|-Score: 0.8133000135421753
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 480
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.1562 - accuracy: 0.6370 - val_loss: 0.7818 - val_accuracy: 0.7388
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6775 - accuracy: 0.7747 - val_loss: 0.6388 - val_accuracy: 0.7872
Trial complete
Trial summary
|-Trial ID: f8a503bef2c79e478007952be17f05b0
|-Score: 0.7871999740600586
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 224
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6631 - accuracy: 0.7677 - val_loss: 0.5159 - val_accuracy: 0.8238
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.4763 - accuracy: 0.8334 - val_loss: 0.4891 - val_accuracy: 0.8251
Trial complete
Trial summary
|-Trial ID: 0157bbe42e294854219bdd3f2ad7edca
|-Score: 0.8251000046730042
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 352
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.2261 - accuracy: 0.6334 - val_loss: 0.8287 - val_accuracy: 0.7345
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7208 - accuracy: 0.7733 - val_loss: 0.6829 - val_accuracy: 0.7780
Trial complete
Trial summary
|-Trial ID: 900ea04227a84067298dd40a0648d82d
|-Score: 0.777999997138977
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 128
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.0711 - accuracy: 0.6752 - val_loss: 0.7499 - val_accuracy: 0.7500
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6554 - accuracy: 0.7841 - val_loss: 0.6243 - val_accuracy: 0.7906
Trial complete
Trial summary
|-Trial ID: 470dda8f3f0e0e18c0fc325d06e66c5e
|-Score: 0.7906000018119812
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 256
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.9834 - accuracy: 0.6994 - val_loss: 0.6892 - val_accuracy: 0.7693
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6139 - accuracy: 0.7997 - val_loss: 0.5998 - val_accuracy: 0.8012
Trial complete
Trial summary
|-Trial ID: 525c300c4de289bc803b85e442558cec
|-Score: 0.8011999726295471
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 384
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6898 - accuracy: 0.7570 - val_loss: 0.5612 - val_accuracy: 0.8008
Epoch 2/2
313/313 [==============================] - 1s 3ms/step - loss: 0.4864 - accuracy: 0.8265 - val_loss: 0.5942 - val_accuracy: 0.7748
Trial complete
Trial summary
|-Trial ID: 93b37da18cbabf429c40385c7d233388
|-Score: 0.8008000254631042
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 160
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7343 - accuracy: 0.7573 - val_loss: 0.6037 - val_accuracy: 0.7778
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.5423 - accuracy: 0.8093 - val_loss: 0.5596 - val_accuracy: 0.7927
Trial complete
Trial summary
|-Trial ID: 3d0626068ddc15eb167d25c7f0e6ef7e
|-Score: 0.7926999926567078
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 352
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 1.1669 - accuracy: 0.6540 - val_loss: 0.8119 - val_accuracy: 0.7407
Epoch 2/2
313/313 [==============================] - 1s 4ms/step - loss: 0.7058 - accuracy: 0.7753 - val_loss: 0.6672 - val_accuracy: 0.7808
Trial complete
Trial summary
|-Trial ID: a3241c44ed4d9f2ace1cc832353e9afb
|-Score: 0.7807999849319458
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 160
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6650 - accuracy: 0.7699 - val_loss: 0.5608 - val_accuracy: 0.8004
Epoch 2/2
313/313 [==============================] - 1s 3ms/step - loss: 0.4872 - accuracy: 0.8298 - val_loss: 0.5115 - val_accuracy: 0.8101
Trial complete
Trial summary
|-Trial ID: e3e8d371362ed1c2b084a0bdbf6e2d04
|-Score: 0.8101000189781189
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 224
Epoch 1/2
313/313 [==============================] - 1s 4ms/step - loss: 0.6889 - accuracy: 0.7511 - val_loss: 0.5375 - val_accuracy: 0.8055
Epoch 2/2
313/313 [==============================] - 1s 3ms/step - loss: 0.5189 - accuracy: 0.8144 - val_loss: 0.5331 - val_accuracy: 0.8059
Trial complete
Trial summary
|-Trial ID: 328929ef7dda3e9582e02e3c49acf38d
|-Score: 0.805899977684021
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 2
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 96
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6526 - accuracy: 0.7698 - val_loss: 0.5390 - val_accuracy: 0.8107
Trial complete
Trial summary
|-Trial ID: a688d4732ab16801845376fbe7394bb3
|-Score: 0.810699999332428
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 6180fcc92dc041b82ec3e4f2ade58b12
|-units: 384
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6942 - accuracy: 0.7579 - val_loss: 0.5139 - val_accuracy: 0.8246
Trial complete
Trial summary
|-Trial ID: c23119a6cb9539a42b6e73294f3a1b9c
|-Score: 0.8245999813079834
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 47f427b9ff04f357806160ab4ee25b81
|-units: 192
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6548 - accuracy: 0.7709 - val_loss: 0.5665 - val_accuracy: 0.7952
Trial complete
Trial summary
|-Trial ID: 0d08cf5cda54aa0f4be2a0da9d9eb14e
|-Score: 0.795199990272522
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 0157bbe42e294854219bdd3f2ad7edca
|-units: 352
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6945 - accuracy: 0.7549 - val_loss: 0.5198 - val_accuracy: 0.8199
Trial complete
Trial summary
|-Trial ID: 8e1fd90d359d3c7ba63959aed0774e7a
|-Score: 0.8198999762535095
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 4bc37ea30b559871e97cb13fb9884507
|-units: 320
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7627 - accuracy: 0.7414 - val_loss: 0.5367 - val_accuracy: 0.8146
Trial complete
Trial summary
|-Trial ID: b7548bcd8cf109a8028e8ea88357da19
|-Score: 0.8145999908447266
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 302193218a5e1690cfe753f1603f1737
|-units: 320
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7055 - accuracy: 0.7506 - val_loss: 0.5869 - val_accuracy: 0.7851
Trial complete
Trial summary
|-Trial ID: f0fe41cc3c1743912ae544d4e02f306a
|-Score: 0.785099983215332
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: cd1ced6111b608f8bf8dd92d1362ade9
|-units: 64
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.8089 - accuracy: 0.7431 - val_loss: 0.6329 - val_accuracy: 0.7731
Trial complete
Trial summary
|-Trial ID: c8edb01cbe1e157567d6d81dfe9fe747
|-Score: 0.7731000185012817
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: c24688f868fcd6c5056e05846148bb68
|-units: 480
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6721 - accuracy: 0.7674 - val_loss: 0.5857 - val_accuracy: 0.7977
Trial complete
Trial summary
|-Trial ID: 1911b7298427680b1a922b86772a0d0d
|-Score: 0.7976999878883362
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: e3e8d371362ed1c2b084a0bdbf6e2d04
|-units: 224
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7268 - accuracy: 0.7505 - val_loss: 0.5805 - val_accuracy: 0.7991
Trial complete
Trial summary
|-Trial ID: 34608bc79ac38d6a22b03c813bd86324
|-Score: 0.7990999817848206
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 328929ef7dda3e9582e02e3c49acf38d
|-units: 96
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 1.0125 - accuracy: 0.6846 - val_loss: 0.7097 - val_accuracy: 0.7659
Trial complete
Trial summary
|-Trial ID: af6097da522316e50affa130c3d65fe0
|-Score: 0.7659000158309937
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 4
|-tuner/epochs: 3
|-tuner/initial_epoch: 2
|-tuner/round: 1
|-tuner/trial_id: 525c300c4de289bc803b85e442558cec
|-units: 384
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6919 - accuracy: 0.7602 - val_loss: 0.5218 - val_accuracy: 0.8180
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4836 - accuracy: 0.8316 - val_loss: 0.4993 - val_accuracy: 0.8294
Trial complete
Trial summary
|-Trial ID: fd3826d96d5ab3d23eb48043b9b1caf8
|-Score: 0.8294000029563904
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 2
|-tuner/trial_id: c23119a6cb9539a42b6e73294f3a1b9c
|-units: 192
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6597 - accuracy: 0.7657 - val_loss: 0.5802 - val_accuracy: 0.7920
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4699 - accuracy: 0.8348 - val_loss: 0.4988 - val_accuracy: 0.8245
Trial complete
Trial summary
|-Trial ID: 2cefdcb1028a370e5a05e95d76e86b67
|-Score: 0.8245000243186951
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 2
|-tuner/trial_id: 8e1fd90d359d3c7ba63959aed0774e7a
|-units: 320
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7513 - accuracy: 0.7504 - val_loss: 0.7005 - val_accuracy: 0.7554
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5099 - accuracy: 0.8167 - val_loss: 0.4975 - val_accuracy: 0.8224
Trial complete
Trial summary
|-Trial ID: 2d6b090a410fa9e1273d2023911e0566
|-Score: 0.8223999738693237
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 2
|-tuner/trial_id: b7548bcd8cf109a8028e8ea88357da19
|-units: 320
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6794 - accuracy: 0.7633 - val_loss: 0.5226 - val_accuracy: 0.8154
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4723 - accuracy: 0.8312 - val_loss: 0.4876 - val_accuracy: 0.8283
Trial complete
Trial summary
|-Trial ID: e27cc23e38337fb1e5c91b7e164b6069
|-Score: 0.8282999992370605
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 2
|-tuner/trial_id: a688d4732ab16801845376fbe7394bb3
|-units: 384
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7459 - accuracy: 0.7386 - val_loss: 0.5555 - val_accuracy: 0.8032
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5012 - accuracy: 0.8195 - val_loss: 0.4997 - val_accuracy: 0.8245
Trial complete
Trial summary
|-Trial ID: c4fc33b7eddf485716f287da1f301f7d
|-Score: 0.8245000243186951
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 4
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 2
|-tuner/trial_id: 34608bc79ac38d6a22b03c813bd86324
|-units: 96
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6905 - accuracy: 0.7645 - val_loss: 0.5681 - val_accuracy: 0.8026
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4846 - accuracy: 0.8281 - val_loss: 0.5003 - val_accuracy: 0.8240
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4298 - accuracy: 0.8438 - val_loss: 0.5062 - val_accuracy: 0.8195
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3958 - accuracy: 0.8600 - val_loss: 0.5176 - val_accuracy: 0.8134
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3717 - accuracy: 0.8654 - val_loss: 0.4378 - val_accuracy: 0.8456
Trial complete
Trial summary
|-Trial ID: 4440f48928327ba6b837b15cc2e520a4
|-Score: 0.8456000089645386
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 3
|-tuner/trial_id: fd3826d96d5ab3d23eb48043b9b1caf8
|-units: 192
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6667 - accuracy: 0.7669 - val_loss: 0.5342 - val_accuracy: 0.8103
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4690 - accuracy: 0.8330 - val_loss: 0.4788 - val_accuracy: 0.8326
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4022 - accuracy: 0.8566 - val_loss: 0.4658 - val_accuracy: 0.8352
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3661 - accuracy: 0.8679 - val_loss: 0.4591 - val_accuracy: 0.8369
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3425 - accuracy: 0.8754 - val_loss: 0.4422 - val_accuracy: 0.8339
Trial complete
Trial summary
|-Trial ID: 69c8fee44e0458d2374e5f216ea673c3
|-Score: 0.836899995803833
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 3
|-tuner/trial_id: e27cc23e38337fb1e5c91b7e164b6069
|-units: 384
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6679 - accuracy: 0.7683 - val_loss: 0.5618 - val_accuracy: 0.7970
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4752 - accuracy: 0.8303 - val_loss: 0.4984 - val_accuracy: 0.8277
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4195 - accuracy: 0.8497 - val_loss: 0.5090 - val_accuracy: 0.8216
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3827 - accuracy: 0.8612 - val_loss: 0.4372 - val_accuracy: 0.8483
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3505 - accuracy: 0.8739 - val_loss: 0.4483 - val_accuracy: 0.8418
Trial complete
Trial summary
|-Trial ID: 17c95f1a6d68ec982da6f3418c8a0a0d
|-Score: 0.8482999801635742
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 3
|-tuner/trial_id: 2cefdcb1028a370e5a05e95d76e86b67
|-units: 320
Epoch 11/20
313/313 [==============================] - 1s 4ms/step - loss: 0.6545 - accuracy: 0.7687 - val_loss: 0.5523 - val_accuracy: 0.8032
Epoch 12/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4580 - accuracy: 0.8365 - val_loss: 0.5480 - val_accuracy: 0.7962
Epoch 13/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4139 - accuracy: 0.8509 - val_loss: 0.4656 - val_accuracy: 0.8310
Epoch 14/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3636 - accuracy: 0.8693 - val_loss: 0.4449 - val_accuracy: 0.8419
Epoch 15/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3498 - accuracy: 0.8716 - val_loss: 0.4588 - val_accuracy: 0.8331
Epoch 16/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3186 - accuracy: 0.8846 - val_loss: 0.4494 - val_accuracy: 0.8398
Epoch 17/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3021 - accuracy: 0.8936 - val_loss: 0.4308 - val_accuracy: 0.8489
Epoch 18/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2788 - accuracy: 0.8977 - val_loss: 0.4160 - val_accuracy: 0.8580
Epoch 19/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2693 - accuracy: 0.8994 - val_loss: 0.4137 - val_accuracy: 0.8559
Epoch 20/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2545 - accuracy: 0.9033 - val_loss: 0.4415 - val_accuracy: 0.8468
Trial complete
Trial summary
|-Trial ID: d81725044b0dfa54b322a65a810f02b7
|-Score: 0.8579999804496765
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 20
|-tuner/initial_epoch: 10
|-tuner/round: 4
|-tuner/trial_id: 17c95f1a6d68ec982da6f3418c8a0a0d
|-units: 320
Epoch 11/20
313/313 [==============================] - 1s 4ms/step - loss: 0.6965 - accuracy: 0.7571 - val_loss: 0.5478 - val_accuracy: 0.8045
Epoch 12/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4815 - accuracy: 0.8311 - val_loss: 0.5044 - val_accuracy: 0.8236
Epoch 13/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4265 - accuracy: 0.8482 - val_loss: 0.4605 - val_accuracy: 0.8357
Epoch 14/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3789 - accuracy: 0.8638 - val_loss: 0.4688 - val_accuracy: 0.8354
Epoch 15/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3596 - accuracy: 0.8724 - val_loss: 0.4856 - val_accuracy: 0.8228
Epoch 16/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3354 - accuracy: 0.8775 - val_loss: 0.4752 - val_accuracy: 0.8332
Epoch 17/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3145 - accuracy: 0.8856 - val_loss: 0.4234 - val_accuracy: 0.8519
Epoch 18/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3000 - accuracy: 0.8910 - val_loss: 0.4274 - val_accuracy: 0.8487
Epoch 19/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2817 - accuracy: 0.8971 - val_loss: 0.4126 - val_accuracy: 0.8545
Epoch 20/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2732 - accuracy: 0.9013 - val_loss: 0.4415 - val_accuracy: 0.8488
Trial complete
Trial summary
|-Trial ID: 99280555edc74cd51ce382df1b2b9dd2
|-Score: 0.8544999957084656
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 4
|-tuner/epochs: 20
|-tuner/initial_epoch: 10
|-tuner/round: 4
|-tuner/trial_id: 4440f48928327ba6b837b15cc2e520a4
|-units: 192
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7570 - accuracy: 0.7500 - val_loss: 0.6200 - val_accuracy: 0.7638
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5045 - accuracy: 0.8202 - val_loss: 0.5067 - val_accuracy: 0.8221
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4520 - accuracy: 0.8364 - val_loss: 0.5734 - val_accuracy: 0.8000
Trial complete
Trial summary
|-Trial ID: ceb3e1602a6542603261877bc9464cc8
|-Score: 0.8220999836921692
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 288
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 1.0443 - accuracy: 0.6696 - val_loss: 0.7189 - val_accuracy: 0.7607
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6388 - accuracy: 0.7905 - val_loss: 0.6172 - val_accuracy: 0.7909
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5543 - accuracy: 0.8199 - val_loss: 0.5722 - val_accuracy: 0.8053
Trial complete
Trial summary
|-Trial ID: df86b0506e582cfd29efc763de8f8247
|-Score: 0.8052999973297119
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 320
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6613 - accuracy: 0.7716 - val_loss: 0.6536 - val_accuracy: 0.7694
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4917 - accuracy: 0.8238 - val_loss: 0.4879 - val_accuracy: 0.8303
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4214 - accuracy: 0.8493 - val_loss: 0.4537 - val_accuracy: 0.8390
Trial complete
Trial summary
|-Trial ID: 58ffcecdde9f800364b17fafba170cf2
|-Score: 0.8389999866485596
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 256
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.8555 - accuracy: 0.7138 - val_loss: 0.6373 - val_accuracy: 0.7760
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5392 - accuracy: 0.8161 - val_loss: 0.5357 - val_accuracy: 0.8138
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4753 - accuracy: 0.8358 - val_loss: 0.5518 - val_accuracy: 0.8006
Trial complete
Trial summary
|-Trial ID: 5c89f39310d0592431174f141f7c0da6
|-Score: 0.8137999773025513
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 32
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6925 - accuracy: 0.7596 - val_loss: 0.5574 - val_accuracy: 0.8101
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4779 - accuracy: 0.8301 - val_loss: 0.5116 - val_accuracy: 0.8240
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4260 - accuracy: 0.8506 - val_loss: 0.4910 - val_accuracy: 0.8284
Trial complete
Trial summary
|-Trial ID: 3229f50d437fe0e7e5b751f7cc88ada1
|-Score: 0.8284000158309937
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 128
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7614 - accuracy: 0.7415 - val_loss: 0.5805 - val_accuracy: 0.8063
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5074 - accuracy: 0.8251 - val_loss: 0.5110 - val_accuracy: 0.8194
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4524 - accuracy: 0.8435 - val_loss: 0.4975 - val_accuracy: 0.8267
Trial complete
Trial summary
|-Trial ID: f612db527fb91cf98918f1f691d0e549
|-Score: 0.82669997215271
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 64
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 1.1429 - accuracy: 0.6426 - val_loss: 0.7873 - val_accuracy: 0.7296
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6877 - accuracy: 0.7775 - val_loss: 0.6605 - val_accuracy: 0.7771
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5903 - accuracy: 0.8072 - val_loss: 0.5990 - val_accuracy: 0.7957
Trial complete
Trial summary
|-Trial ID: f767283675e3f4c0c08c7ce5625b6cd7
|-Score: 0.7957000136375427
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 192
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7022 - accuracy: 0.7510 - val_loss: 0.5567 - val_accuracy: 0.7944
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5072 - accuracy: 0.8153 - val_loss: 0.5024 - val_accuracy: 0.8242
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4508 - accuracy: 0.8384 - val_loss: 0.5321 - val_accuracy: 0.8094
Trial complete
Trial summary
|-Trial ID: 9fb3916a9e5d1ed009780203ed464119
|-Score: 0.8241999745368958
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 128
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7197 - accuracy: 0.7473 - val_loss: 0.5333 - val_accuracy: 0.8114
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4953 - accuracy: 0.8234 - val_loss: 0.5239 - val_accuracy: 0.8154
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4626 - accuracy: 0.8308 - val_loss: 0.5280 - val_accuracy: 0.8131
Trial complete
Trial summary
|-Trial ID: 4a7165940cb1a240ed2565e4261b1a3a
|-Score: 0.8154000043869019
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 192
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.7331 - accuracy: 0.7464 - val_loss: 0.6058 - val_accuracy: 0.7704
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5029 - accuracy: 0.8181 - val_loss: 0.4971 - val_accuracy: 0.8196
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4522 - accuracy: 0.8381 - val_loss: 0.4903 - val_accuracy: 0.8314
Trial complete
Trial summary
|-Trial ID: f8cbfe40d608cf705017373eb66a4e48
|-Score: 0.8313999772071838
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 256
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6457 - accuracy: 0.7756 - val_loss: 0.5157 - val_accuracy: 0.8202
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4635 - accuracy: 0.8302 - val_loss: 0.4819 - val_accuracy: 0.8303
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.4066 - accuracy: 0.8524 - val_loss: 0.4512 - val_accuracy: 0.8442
Trial complete
Trial summary
|-Trial ID: 85ec8cf3b484e5c0a676f4dfbe2ec629
|-Score: 0.8442000150680542
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 416
Epoch 1/3
313/313 [==============================] - 1s 4ms/step - loss: 0.9563 - accuracy: 0.7019 - val_loss: 0.6858 - val_accuracy: 0.7605
Epoch 2/3
313/313 [==============================] - 1s 4ms/step - loss: 0.6072 - accuracy: 0.8002 - val_loss: 0.5848 - val_accuracy: 0.8023
Epoch 3/3
313/313 [==============================] - 1s 4ms/step - loss: 0.5275 - accuracy: 0.8252 - val_loss: 0.5461 - val_accuracy: 0.8129
Trial complete
Trial summary
|-Trial ID: b4fcce3ed3fae99880e3a85c3941b7ac
|-Score: 0.8129000067710876
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 3
|-tuner/epochs: 3
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 480
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6547 - accuracy: 0.7719 - val_loss: 0.5493 - val_accuracy: 0.8090
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4669 - accuracy: 0.8360 - val_loss: 0.4915 - val_accuracy: 0.8256
Trial complete
Trial summary
|-Trial ID: 2fac18ea504beafe35bfa988c35280f9
|-Score: 0.8256000280380249
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: 85ec8cf3b484e5c0a676f4dfbe2ec629
|-units: 416
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6743 - accuracy: 0.7633 - val_loss: 0.6164 - val_accuracy: 0.7704
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4788 - accuracy: 0.8284 - val_loss: 0.4915 - val_accuracy: 0.8294
Trial complete
Trial summary
|-Trial ID: 598ba594fcd70be025263ece46f42e0a
|-Score: 0.8294000029563904
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: 58ffcecdde9f800364b17fafba170cf2
|-units: 256
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7133 - accuracy: 0.7530 - val_loss: 0.6553 - val_accuracy: 0.7702
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5097 - accuracy: 0.8168 - val_loss: 0.4828 - val_accuracy: 0.8337
Trial complete
Trial summary
|-Trial ID: b8f0b0b73861da5e37fca6c27a46500e
|-Score: 0.8337000012397766
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: f8cbfe40d608cf705017373eb66a4e48
|-units: 256
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7133 - accuracy: 0.7538 - val_loss: 0.5637 - val_accuracy: 0.8000
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4909 - accuracy: 0.8300 - val_loss: 0.5109 - val_accuracy: 0.8213
Trial complete
Trial summary
|-Trial ID: 31ae14872b224ce0fcb35eab028ea6ca
|-Score: 0.8213000297546387
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: 3229f50d437fe0e7e5b751f7cc88ada1
|-units: 128
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7929 - accuracy: 0.7338 - val_loss: 0.5711 - val_accuracy: 0.8028
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5078 - accuracy: 0.8252 - val_loss: 0.5119 - val_accuracy: 0.8211
Trial complete
Trial summary
|-Trial ID: 9f32f6942ae4172266e73ddb9fe1a0b1
|-Score: 0.8210999965667725
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: f612db527fb91cf98918f1f691d0e549
|-units: 64
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6955 - accuracy: 0.7454 - val_loss: 0.5868 - val_accuracy: 0.7754
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5016 - accuracy: 0.8188 - val_loss: 0.4745 - val_accuracy: 0.8288
Trial complete
Trial summary
|-Trial ID: 0a5d8689f797941b75813239c3b389ff
|-Score: 0.8288000226020813
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 5
|-tuner/initial_epoch: 3
|-tuner/round: 1
|-tuner/trial_id: 9fb3916a9e5d1ed009780203ed464119
|-units: 128
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7253 - accuracy: 0.7495 - val_loss: 0.5184 - val_accuracy: 0.8203
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5220 - accuracy: 0.8107 - val_loss: 0.5751 - val_accuracy: 0.8020
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4602 - accuracy: 0.8342 - val_loss: 0.5813 - val_accuracy: 0.7992
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4232 - accuracy: 0.8460 - val_loss: 0.5204 - val_accuracy: 0.8218
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4193 - accuracy: 0.8475 - val_loss: 0.4857 - val_accuracy: 0.8357
Trial complete
Trial summary
|-Trial ID: 2d6a93aaa3e9a4d83f78e014e5e2b737
|-Score: 0.8356999754905701
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 2
|-tuner/trial_id: b8f0b0b73861da5e37fca6c27a46500e
|-units: 256
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6880 - accuracy: 0.7618 - val_loss: 0.5582 - val_accuracy: 0.8071
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4774 - accuracy: 0.8365 - val_loss: 0.5277 - val_accuracy: 0.8167
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4222 - accuracy: 0.8502 - val_loss: 0.4509 - val_accuracy: 0.8420
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3752 - accuracy: 0.8648 - val_loss: 0.4962 - val_accuracy: 0.8221
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3468 - accuracy: 0.8780 - val_loss: 0.4366 - val_accuracy: 0.8473
Trial complete
Trial summary
|-Trial ID: 655be1106a54585d35d8f5ef134795a8
|-Score: 0.8472999930381775
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 2
|-tuner/trial_id: 598ba594fcd70be025263ece46f42e0a
|-units: 256
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7173 - accuracy: 0.7449 - val_loss: 0.5442 - val_accuracy: 0.8110
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4938 - accuracy: 0.8216 - val_loss: 0.5118 - val_accuracy: 0.8226
Epoch 8/10
313/313 [==============================] - 1s 3ms/step - loss: 0.4529 - accuracy: 0.8344 - val_loss: 0.5303 - val_accuracy: 0.8124
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4232 - accuracy: 0.8456 - val_loss: 0.4821 - val_accuracy: 0.8323
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4108 - accuracy: 0.8483 - val_loss: 0.4961 - val_accuracy: 0.8309
Trial complete
Trial summary
|-Trial ID: 3fc067f392e3f1e54ccae61b78b112a7
|-Score: 0.8323000073432922
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 2
|-tuner/trial_id: 0a5d8689f797941b75813239c3b389ff
|-units: 128
Epoch 11/20
313/313 [==============================] - 1s 4ms/step - loss: 0.6745 - accuracy: 0.7660 - val_loss: 0.6225 - val_accuracy: 0.7813
Epoch 12/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4846 - accuracy: 0.8342 - val_loss: 0.4840 - val_accuracy: 0.8328
Epoch 13/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4138 - accuracy: 0.8523 - val_loss: 0.4514 - val_accuracy: 0.8428
Epoch 14/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3832 - accuracy: 0.8619 - val_loss: 0.4588 - val_accuracy: 0.8393
Epoch 15/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3552 - accuracy: 0.8734 - val_loss: 0.5335 - val_accuracy: 0.8010
Epoch 16/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3252 - accuracy: 0.8840 - val_loss: 0.4203 - val_accuracy: 0.8467
Epoch 17/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3100 - accuracy: 0.8862 - val_loss: 0.4558 - val_accuracy: 0.8485
Epoch 18/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2892 - accuracy: 0.8946 - val_loss: 0.4903 - val_accuracy: 0.8325
Epoch 19/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2809 - accuracy: 0.8950 - val_loss: 0.4734 - val_accuracy: 0.8380
Epoch 20/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2596 - accuracy: 0.9048 - val_loss: 0.4197 - val_accuracy: 0.8562
Trial complete
Trial summary
|-Trial ID: 1f44a392dde68f54886e7da19ca65448
|-Score: 0.8561999797821045
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 3
|-tuner/epochs: 20
|-tuner/initial_epoch: 10
|-tuner/round: 3
|-tuner/trial_id: 655be1106a54585d35d8f5ef134795a8
|-units: 256
Epoch 11/20
313/313 [==============================] - 1s 4ms/step - loss: 0.7455 - accuracy: 0.7452 - val_loss: 0.5645 - val_accuracy: 0.7993
Epoch 12/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4985 - accuracy: 0.8185 - val_loss: 0.5247 - val_accuracy: 0.8107
Epoch 13/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4568 - accuracy: 0.8361 - val_loss: 0.5267 - val_accuracy: 0.8131
Epoch 14/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4254 - accuracy: 0.8429 - val_loss: 0.4992 - val_accuracy: 0.8279
Epoch 15/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4019 - accuracy: 0.8528 - val_loss: 0.4895 - val_accuracy: 0.8306
Epoch 16/20
313/313 [==============================] - 1s 3ms/step - loss: 0.4059 - accuracy: 0.8521 - val_loss: 0.5547 - val_accuracy: 0.7995
Epoch 17/20
313/313 [==============================] - 1s 3ms/step - loss: 0.3927 - accuracy: 0.8541 - val_loss: 0.5153 - val_accuracy: 0.8280
Epoch 18/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3667 - accuracy: 0.8656 - val_loss: 0.5395 - val_accuracy: 0.8187
Epoch 19/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3501 - accuracy: 0.8696 - val_loss: 0.4756 - val_accuracy: 0.8395
Epoch 20/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3424 - accuracy: 0.8703 - val_loss: 0.5030 - val_accuracy: 0.8324
Trial complete
Trial summary
|-Trial ID: 30df8ce66b0d8785077b0f12619880d5
|-Score: 0.8395000100135803
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 3
|-tuner/epochs: 20
|-tuner/initial_epoch: 10
|-tuner/round: 3
|-tuner/trial_id: 2d6a93aaa3e9a4d83f78e014e5e2b737
|-units: 256
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7179 - accuracy: 0.7512 - val_loss: 0.5625 - val_accuracy: 0.7825
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4952 - accuracy: 0.8153 - val_loss: 0.5170 - val_accuracy: 0.8156
Epoch 3/5
313/313 [==============================] - 1s 3ms/step - loss: 0.4424 - accuracy: 0.8390 - val_loss: 0.4988 - val_accuracy: 0.8138
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4260 - accuracy: 0.8437 - val_loss: 0.5157 - val_accuracy: 0.8178
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.3967 - accuracy: 0.8519 - val_loss: 0.6059 - val_accuracy: 0.7844
Trial complete
Trial summary
|-Trial ID: c8e966b396e210c35657ae69f9adeaab
|-Score: 0.817799985408783
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 224
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7269 - accuracy: 0.7468 - val_loss: 0.5776 - val_accuracy: 0.7907
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4841 - accuracy: 0.8263 - val_loss: 0.5111 - val_accuracy: 0.8235
Epoch 3/5
313/313 [==============================] - 1s 3ms/step - loss: 0.4430 - accuracy: 0.8426 - val_loss: 0.5254 - val_accuracy: 0.8054
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4006 - accuracy: 0.8611 - val_loss: 0.4650 - val_accuracy: 0.8374
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.3739 - accuracy: 0.8656 - val_loss: 0.4442 - val_accuracy: 0.8439
Trial complete
Trial summary
|-Trial ID: 0cbc50970c0a13b3f14aabe114e37684
|-Score: 0.8439000248908997
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 96
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.7813 - accuracy: 0.7503 - val_loss: 0.5447 - val_accuracy: 0.8050
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5314 - accuracy: 0.8120 - val_loss: 0.5416 - val_accuracy: 0.8059
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4926 - accuracy: 0.8192 - val_loss: 0.5307 - val_accuracy: 0.8258
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4544 - accuracy: 0.8377 - val_loss: 0.5648 - val_accuracy: 0.8062
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4117 - accuracy: 0.8499 - val_loss: 0.4996 - val_accuracy: 0.8296
Trial complete
Trial summary
|-Trial ID: 1394b38972c3aab7b04eeb51d426d6bd
|-Score: 0.8295999765396118
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 512
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 1.0113 - accuracy: 0.6870 - val_loss: 0.7039 - val_accuracy: 0.7684
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6092 - accuracy: 0.7983 - val_loss: 0.5935 - val_accuracy: 0.7984
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5284 - accuracy: 0.8260 - val_loss: 0.5540 - val_accuracy: 0.8113
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4852 - accuracy: 0.8370 - val_loss: 0.5185 - val_accuracy: 0.8241
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4581 - accuracy: 0.8463 - val_loss: 0.4975 - val_accuracy: 0.8307
Trial complete
Trial summary
|-Trial ID: 73e06a845b8b449954109979013ad802
|-Score: 0.8306999802589417
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 416
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.9611 - accuracy: 0.6970 - val_loss: 0.6773 - val_accuracy: 0.7801
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6034 - accuracy: 0.8020 - val_loss: 0.6036 - val_accuracy: 0.7908
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5247 - accuracy: 0.8264 - val_loss: 0.5601 - val_accuracy: 0.8106
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4811 - accuracy: 0.8387 - val_loss: 0.5183 - val_accuracy: 0.8240
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4527 - accuracy: 0.8488 - val_loss: 0.5007 - val_accuracy: 0.8272
Trial complete
Trial summary
|-Trial ID: dd30179d4cb64d4a5993f36ca78e44b6
|-Score: 0.8271999955177307
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 512
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6848 - accuracy: 0.7574 - val_loss: 0.5529 - val_accuracy: 0.7988
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5228 - accuracy: 0.8109 - val_loss: 0.5548 - val_accuracy: 0.7938
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4635 - accuracy: 0.8369 - val_loss: 0.4992 - val_accuracy: 0.8224
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4427 - accuracy: 0.8418 - val_loss: 0.5387 - val_accuracy: 0.8049
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4141 - accuracy: 0.8452 - val_loss: 0.5447 - val_accuracy: 0.8104
Trial complete
Trial summary
|-Trial ID: 00c4220044fc95e5d54a8612f8bbc35a
|-Score: 0.8223999738693237
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 32
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.9881 - accuracy: 0.6954 - val_loss: 0.7038 - val_accuracy: 0.7621
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6110 - accuracy: 0.7990 - val_loss: 0.5933 - val_accuracy: 0.8007
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.5327 - accuracy: 0.8233 - val_loss: 0.5433 - val_accuracy: 0.8129
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4863 - accuracy: 0.8385 - val_loss: 0.5602 - val_accuracy: 0.7996
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4560 - accuracy: 0.8466 - val_loss: 0.5032 - val_accuracy: 0.8246
Trial complete
Trial summary
|-Trial ID: 1ba7b042423be6317ecb4f2782aa68c6
|-Score: 0.8245999813079834
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 448
Epoch 1/5
313/313 [==============================] - 1s 4ms/step - loss: 0.6564 - accuracy: 0.7715 - val_loss: 0.6531 - val_accuracy: 0.7719
Epoch 2/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4629 - accuracy: 0.8344 - val_loss: 0.5060 - val_accuracy: 0.8235
Epoch 3/5
313/313 [==============================] - 1s 4ms/step - loss: 0.4174 - accuracy: 0.8489 - val_loss: 0.4695 - val_accuracy: 0.8356
Epoch 4/5
313/313 [==============================] - 1s 4ms/step - loss: 0.3647 - accuracy: 0.8687 - val_loss: 0.4558 - val_accuracy: 0.8386
Epoch 5/5
313/313 [==============================] - 1s 4ms/step - loss: 0.3468 - accuracy: 0.8727 - val_loss: 0.4561 - val_accuracy: 0.8399
Trial complete
Trial summary
|-Trial ID: e723dee274f99c625bb6002f38e8a5a0
|-Score: 0.839900016784668
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 5
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 480
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7285 - accuracy: 0.7491 - val_loss: 0.5759 - val_accuracy: 0.8020
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5086 - accuracy: 0.8209 - val_loss: 0.5622 - val_accuracy: 0.7995
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4502 - accuracy: 0.8426 - val_loss: 0.5000 - val_accuracy: 0.8293
Epoch 9/10
313/313 [==============================] - 1s 3ms/step - loss: 0.4162 - accuracy: 0.8558 - val_loss: 0.4584 - val_accuracy: 0.8422
Epoch 10/10
313/313 [==============================] - 1s 3ms/step - loss: 0.3897 - accuracy: 0.8612 - val_loss: 0.4605 - val_accuracy: 0.8398
Trial complete
Trial summary
|-Trial ID: 2be62d66a3e254d473df0eb4838f3a03
|-Score: 0.842199981212616
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 1
|-tuner/trial_id: 0cbc50970c0a13b3f14aabe114e37684
|-units: 96
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6614 - accuracy: 0.7685 - val_loss: 0.5806 - val_accuracy: 0.7835
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4726 - accuracy: 0.8363 - val_loss: 0.4760 - val_accuracy: 0.8330
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4027 - accuracy: 0.8545 - val_loss: 0.4767 - val_accuracy: 0.8335
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3684 - accuracy: 0.8676 - val_loss: 0.4380 - val_accuracy: 0.8448
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3404 - accuracy: 0.8752 - val_loss: 0.4306 - val_accuracy: 0.8493
Trial complete
Trial summary
|-Trial ID: fd05cafeccb89704c68a8f521f9dae7a
|-Score: 0.8493000268936157
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 1
|-tuner/trial_id: e723dee274f99c625bb6002f38e8a5a0
|-units: 480
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 1.0227 - accuracy: 0.6808 - val_loss: 0.7094 - val_accuracy: 0.7645
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.6159 - accuracy: 0.7966 - val_loss: 0.5948 - val_accuracy: 0.7969
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5342 - accuracy: 0.8218 - val_loss: 0.5441 - val_accuracy: 0.8185
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4924 - accuracy: 0.8343 - val_loss: 0.5255 - val_accuracy: 0.8196
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4586 - accuracy: 0.8458 - val_loss: 0.5054 - val_accuracy: 0.8247
Trial complete
Trial summary
|-Trial ID: f8d037048c258d8a78c0546d0001ac69
|-Score: 0.8246999979019165
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.0001
|-tuner/bracket: 2
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 1
|-tuner/trial_id: 73e06a845b8b449954109979013ad802
|-units: 416
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7788 - accuracy: 0.7472 - val_loss: 0.5943 - val_accuracy: 0.7932
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5115 - accuracy: 0.8176 - val_loss: 0.5834 - val_accuracy: 0.7933
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4541 - accuracy: 0.8322 - val_loss: 0.5334 - val_accuracy: 0.8073
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4219 - accuracy: 0.8477 - val_loss: 0.5085 - val_accuracy: 0.8116
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4093 - accuracy: 0.8503 - val_loss: 0.5157 - val_accuracy: 0.8334
Trial complete
Trial summary
|-Trial ID: a80ec8308aba53f3c2a8702ec49df6b9
|-Score: 0.8334000110626221
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 2
|-tuner/epochs: 10
|-tuner/initial_epoch: 5
|-tuner/round: 1
|-tuner/trial_id: 1394b38972c3aab7b04eeb51d426d6bd
|-units: 512
Epoch 11/20
313/313 [==============================] - 1s 4ms/step - loss: 0.6539 - accuracy: 0.7685 - val_loss: 0.5531 - val_accuracy: 0.8021
Epoch 12/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4650 - accuracy: 0.8353 - val_loss: 0.5747 - val_accuracy: 0.7955
Epoch 13/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4098 - accuracy: 0.8489 - val_loss: 0.4436 - val_accuracy: 0.8419
Epoch 14/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3766 - accuracy: 0.8638 - val_loss: 0.4500 - val_accuracy: 0.8385
Epoch 15/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3397 - accuracy: 0.8736 - val_loss: 0.4233 - val_accuracy: 0.8459
Epoch 16/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3216 - accuracy: 0.8824 - val_loss: 0.4282 - val_accuracy: 0.8481
Epoch 17/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2894 - accuracy: 0.8918 - val_loss: 0.4659 - val_accuracy: 0.8424
Epoch 18/20
313/313 [==============================] - 1s 3ms/step - loss: 0.2841 - accuracy: 0.8939 - val_loss: 0.4762 - val_accuracy: 0.8356
Epoch 19/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2588 - accuracy: 0.9038 - val_loss: 0.4104 - val_accuracy: 0.8578
Epoch 20/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2571 - accuracy: 0.9068 - val_loss: 0.4156 - val_accuracy: 0.8544
Trial complete
Trial summary
|-Trial ID: 7bfd1254b490b449a9d9a3f4b12925de
|-Score: 0.8578000068664551
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 20
|-tuner/initial_epoch: 10
|-tuner/round: 2
|-tuner/trial_id: fd05cafeccb89704c68a8f521f9dae7a
|-units: 480
Epoch 11/20
313/313 [==============================] - 1s 4ms/step - loss: 0.7175 - accuracy: 0.7521 - val_loss: 0.5640 - val_accuracy: 0.8036
Epoch 12/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4932 - accuracy: 0.8260 - val_loss: 0.5487 - val_accuracy: 0.8074
Epoch 13/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4338 - accuracy: 0.8450 - val_loss: 0.5174 - val_accuracy: 0.8063
Epoch 14/20
313/313 [==============================] - 1s 4ms/step - loss: 0.4061 - accuracy: 0.8564 - val_loss: 0.4603 - val_accuracy: 0.8378
Epoch 15/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3787 - accuracy: 0.8634 - val_loss: 0.5172 - val_accuracy: 0.8237
Epoch 16/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3584 - accuracy: 0.8708 - val_loss: 0.4362 - val_accuracy: 0.8477
Epoch 17/20
313/313 [==============================] - 1s 3ms/step - loss: 0.3326 - accuracy: 0.8826 - val_loss: 0.4464 - val_accuracy: 0.8449
Epoch 18/20
313/313 [==============================] - 1s 4ms/step - loss: 0.3212 - accuracy: 0.8837 - val_loss: 0.4274 - val_accuracy: 0.8494
Epoch 19/20
313/313 [==============================] - 1s 3ms/step - loss: 0.3065 - accuracy: 0.8866 - val_loss: 0.4368 - val_accuracy: 0.8464
Epoch 20/20
313/313 [==============================] - 1s 4ms/step - loss: 0.2906 - accuracy: 0.8935 - val_loss: 0.4210 - val_accuracy: 0.8525
Trial complete
Trial summary
|-Trial ID: aac08287e2b5aeddf7d9f8638819c084
|-Score: 0.8525000214576721
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.001
|-tuner/bracket: 2
|-tuner/epochs: 20
|-tuner/initial_epoch: 10
|-tuner/round: 2
|-tuner/trial_id: 2be62d66a3e254d473df0eb4838f3a03
|-units: 96
Epoch 1/10
313/313 [==============================] - 1s 4ms/step - loss: 0.7491 - accuracy: 0.7533 - val_loss: 0.5922 - val_accuracy: 0.7665
Epoch 2/10
313/313 [==============================] - 1s 4ms/step - loss: 0.5156 - accuracy: 0.8142 - val_loss: 0.5375 - val_accuracy: 0.8178
Epoch 3/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4531 - accuracy: 0.8343 - val_loss: 0.6348 - val_accuracy: 0.7866
Epoch 4/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4311 - accuracy: 0.8409 - val_loss: 0.5099 - val_accuracy: 0.8315
Epoch 5/10
313/313 [==============================] - 1s 4ms/step - loss: 0.4028 - accuracy: 0.8537 - val_loss: 0.4864 - val_accuracy: 0.8356
Epoch 6/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3673 - accuracy: 0.8640 - val_loss: 0.5126 - val_accuracy: 0.8292
Epoch 7/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3631 - accuracy: 0.8625 - val_loss: 0.5416 - val_accuracy: 0.8171
Epoch 8/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3586 - accuracy: 0.8679 - val_loss: 0.4798 - val_accuracy: 0.8414
Epoch 9/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3416 - accuracy: 0.8726 - val_loss: 0.6085 - val_accuracy: 0.8127
Epoch 10/10
313/313 [==============================] - 1s 4ms/step - loss: 0.3612 - accuracy: 0.8674 - val_loss: 0.5086 - val_accuracy: 0.8311
Trial complete
Trial summary
|-Trial ID: 7efd24a64590e14b5dc6156afcf0ce0e
|-Score: 0.8414000272750854
|-Best step: 0
Hyperparameters:
|-learning_rate: 0.01
|-tuner/bracket: 1
|-tuner/epochs: 10
|-tuner/initial_epoch: 0
|-tuner/round: 0
|-units: 384
INFO:tensorflow:Oracle triggered exitbracket4
エポック数：2→3→5→10→20
モデル数：20→10→5→3→2
bracket3
エポック数：3→5→10→20
モデル数：12→6→3→2
bracket2
エポック数：5→10→20
モデル数：8→4→2
bracket1
エポック数：10
モデル数：1
tuner.search()でepochs=10にしたときは
bracket3、モデル数10から探索が始まりましたが、
tuner.search()でepochs=20にした今回は
bracket4、モデル数20から探索が始まりました。
ハイパーパラメータの組み合わせが45通りあるときは
epochs=45
で
tuner.search()
を実行してやれば全モデルに対して最低1回は探索が入りそうです。
Kerasチューナーの紹介
上記のチュートリアルを読んでもepochsの初期値とfactorの話については一切触れられていないのでちょっと不親切かなと思いました。
解読に結構時間がかかった。