-
Notifications
You must be signed in to change notification settings - Fork 1
/
search_best_hparams.py
53 lines (44 loc) · 1.63 KB
/
search_best_hparams.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
import time
import pickle
import tensorflow as tf
from kerastuner.tuners import RandomSearch
from data.preprocessing import prepare_data
from tensorflow.keras.models import Sequential
from sklearn.model_selection import train_test_split
from tensorflow.keras.layers import Dense, LSTM, Embedding
def build_model(hp):
output_dim = hp.Int("output_dim", min_value=32, max_value=512, step=16)
input_unit = hp.Int("input_unit", min_value=32, max_value=512, step=16)
model = Sequential()
model.add(Embedding(input_dim=260383, output_dim=output_dim, input_length=500))
model.add(LSTM(input_unit))
model.add(Dense(5, activation='sigmoid'))
hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
model.compile(loss='categorical_crossentropy',
optimizer=tf.keras.optimizers.Adam(learning_rate=hp_learning_rate),
metrics=['accuracy'])
return model
all_encoded_texts, all_labels, INPUT_DIM = prepare_data(tokenizer_path=None)
X_train, X_test, y_train, y_test = train_test_split(
all_encoded_texts,
all_labels,
test_size=0.2
)
if __name__ == "__main__":
tuner = RandomSearch(
build_model,
objective = "val_accuracy",
max_trials = 20,
executions_per_trial = 3,
directory = f"tuners/trial_model_{int(time.time())}",
project_name = "news_classifier"
)
tuner.search(
x = X_train,
y = y_train,
epochs = 6,
batch_size = 64,
validation_data = (X_test, y_test)
)
with open(f"tuner_main{int(time.time())}.pkl", "wb") as f:
pickle.dump(tuner, f)