commit 71d635f2bb7a89665a0e7f4964accd92393499fc
parent de1869929bb860e6ba00520bd523083d37319e10
Author: MTRNord <mtrnord1@gmail.com>
Date: Wed, 23 Nov 2022 21:31:00 +0100
Fixup model py
Diffstat:
| M | model_v2.py | | | 100 | ++++++++++++++++++++++--------------------------------------------------------- |
1 file changed, 27 insertions(+), 73 deletions(-)
diff --git a/model_v2.py b/model_v2.py
@@ -43,16 +43,8 @@ progress_bar = tf.keras.callbacks.ProgbarLogger()
class SpamDectionModel(tf.keras.Model):
- def __init__(self, vectorize_layer, hp_embedding_dim, hp_dense1, hp_dropout, hp_l2):
+ def __init__(self, hp_dense1, hp_dropout, hp_l2):
super(SpamDectionModel, self).__init__()
- # self.vectorize_layer = vectorize_layer
- # self.embedding = tf.keras.layers.Embedding(
- # input_dim=len(vectorize_layer.get_vocabulary()) + 1,
- # output_dim=hp_embedding_dim,
- # name="text_input",
- # embeddings_initializer="uniform",
- # # mask_zero=True,
- # )
self.dropout = tf.keras.layers.Dropout(
hp_dropout,
)
@@ -75,9 +67,6 @@ class SpamDectionModel(tf.keras.Model):
@tf.function
def call(self, x, training=False):
- # x = self.vectorize_layer(x)
- # x = self.embedding(x)
- # x = self.glob_average_pooling_1d(x)
x = self.use_layer(x)
if training:
x = self.dropout(x, training=training)
@@ -88,22 +77,16 @@ class SpamDectionModel(tf.keras.Model):
class SpamDectionHyperModel(kt.HyperModel):
- def __init__(self, vectorize_layer):
+ def __init__(self):
super(SpamDectionHyperModel, self).__init__()
- self.vectorize_layer = vectorize_layer
def build(self, hp):
# Tune the number of units in the first Dense layer
# Choose an optimal value between 6-512
hp_dense1 = hp.Int("dense1", min_value=6, max_value=512, step=12)
- hp_embedding_dim = hp.Int(
- "embedding_dim", min_value=300, max_value=512, step=16
- )
hp_dropout = hp.Float("dropout", min_value=0.1, max_value=0.9, step=0.1)
hp_l2 = hp.Float("l2", min_value=0.0001, max_value=0.001, step=0.0001)
model = SpamDectionModel(
- self.vectorize_layer,
- hp_embedding_dim + 1,
hp_dense1,
hp_dropout,
hp_l2,
@@ -152,25 +135,6 @@ def remove_stopwords(input_text):
def change_labels(x):
return 1 if x == "spam" else 0
-
-def tokenize_data(data):
- vectorize_layer = tf.keras.layers.TextVectorization(
- max_tokens=vocab_size,
- output_mode="int",
- standardize="lower_and_strip_punctuation",
- )
-
- # Now that the vocab layer has been created, call `adapt` on the text-only
- # dataset to create the vocabulary. You don't have to batch, but for large
- # datasets this means we're not keeping spare copies of the dataset.
- vectorize_layer.adapt(data)
-
- # vocab = np.array(vectorize_layer.get_vocabulary())
- # print(f"First 20 of Vocab: {vocab[:20]}")
-
- return vectorize_layer
-
-
def load_data():
data = pd.read_csv(
"./input/MatrixData.tsv", sep="\t", quoting=csv.QUOTE_NONE, encoding="utf-8"
@@ -206,9 +170,7 @@ def load_data():
training_labels_final = np.array(training_labels)
training_sentences_final = np.array(training_sentences)
testing_sentences_final = np.array(testing_sentences)
- vectorize_layer = tokenize_data(sentences)
return (
- vectorize_layer,
training_sentences_final,
testing_sentences_final,
training_labels_final,
@@ -249,17 +211,14 @@ def train_hyperparamters(
def train_model(
- vectorize_layer,
training_sentences_final,
testing_sentences_final,
training_labels_final,
testing_labels_final,
- # best_hps,
- # tuner,
+ best_hps,
+ tuner,
):
model = SpamDectionModel(
- vectorize_layer,
- 0,
64,
0.2,
0,
@@ -270,7 +229,7 @@ def train_model(
metrics=["accuracy"],
)
num_epochs = 200
- # model = tuner.hypermodel.build(best_hps)
+ model = tuner.hypermodel.build(best_hps)
history = model.fit(
training_sentences_final,
training_labels_final,
@@ -286,8 +245,6 @@ def train_model(
print("Average test loss: ", np.average(history.history["val_loss"]))
model = SpamDectionModel(
- vectorize_layer,
- 0,
64,
0.2,
0,
@@ -297,8 +254,8 @@ def train_model(
optimizer=tf.keras.optimizers.Adam(),
metrics=["accuracy"],
)
- # hypermodel = tuner.hypermodel.build(best_hps)
- hypermodel_history = model.fit(
+ hypermodel = tuner.hypermodel.build(best_hps)
+ hypermodel_history = hypermodel.fit(
training_sentences_final,
training_labels_final,
verbose=1,
@@ -310,7 +267,6 @@ def train_model(
# filepath=checkpoint_prefix, save_weights_only=True
# ),
progress_bar
- # es_callback
],
)
@@ -326,7 +282,7 @@ def train_model(
return model
-def test_model(vectorize_layer, model):
+def test_model(model):
# Use the model to predict whether a message is spam
text_messages = [
"Greg, can you call me back once you get this?",
@@ -400,39 +356,37 @@ def main():
print("TensorFlow version:", tf.__version__)
print("[Step 1/6] Loading data")
(
- vectorize_layer,
training_sentences_final,
testing_sentences_final,
training_labels_final,
testing_labels_final,
) = load_data()
- # model = SpamDectionHyperModel(vectorize_layer)
- # tuner = kt.Hyperband(
- # model,
- # objective="val_accuracy",
- # max_epochs=100,
- # directory="hyper_tuning",
- # project_name="spam-keras",
- # )
- # print("[Step 3/6] Tuning hypervalues")
- # best_hps = train_hyperparamters(
- # training_sentences_final,
- # testing_sentences_final,
- # training_labels_final,
- # testing_labels_final,
- # tuner,
- # )
+ model = SpamDectionHyperModel()
+ tuner = kt.Hyperband(
+ model,
+ objective="val_accuracy",
+ max_epochs=100,
+ directory="hyper_tuning",
+ project_name="spam-keras",
+ )
+ print("[Step 3/6] Tuning hypervalues")
+ best_hps = train_hyperparamters(
+ training_sentences_final,
+ testing_sentences_final,
+ training_labels_final,
+ testing_labels_final,
+ tuner,
+ )
print("[Step 4/6] Training model")
model = train_model(
- vectorize_layer,
training_sentences_final,
testing_sentences_final,
training_labels_final,
testing_labels_final,
- # best_hps,
- # tuner,
+ best_hps,
+ tuner,
)
print("[Step 5/6] Saving model")
@@ -442,7 +396,7 @@ def main():
model.save(export_path)
print("[Step 6/6] Testing model")
- test_model(vectorize_layer, model)
+ test_model(model)
if __name__ == "__main__":
