def main(argv):
args = parser.parse_args(argv[1:])
# Get Feature Columns
feature_columns = data_load.get_feature_column(
"data/original_features.csv")
# Build 2 hidden layer DNN with 20, 20 units respectively.
checkpointing_config = tf.estimator.RunConfig(
save_checkpoints_secs=60, # Save checkpoints every 60 seconds.
keep_checkpoint_max=10, # Retain the 10 most recent checkpoints.
)
classifier = tf.estimator.DNNClassifier(
model_dir="model/y_dnn",
config=checkpointing_config,
feature_columns=feature_columns,
# Two hidden layers of 20 nodes each.
hidden_units=[20, 20],
# The model must choose between 2 classes.
n_classes=2)
train_x, train_y = get_data_to_be_re_train()
# Train the Model.
classifier.train(
input_fn=lambda: train_input_fn(train_x, train_y, args.batch_size),
steps=args.train_steps)
def main(argv):
args = parser.parse_args(argv[1:])
# Get Feature Columns
feature_columns = data_load.get_feature_column("data/original_features.csv")
# Build 2 hidden layer DNN with 20, 20 units respectively.
checkpointing_config = tf.estimator.RunConfig(
save_checkpoints_secs=60, # Save checkpoints every 60 seconds.
keep_checkpoint_max=10, # Retain the 10 most recent checkpoints.
)
classifier = tf.estimator.DNNClassifier(
model_dir="model/y_dnn",
config=checkpointing_config,
feature_columns=feature_columns,
# Two hidden layers of 20 nodes each.
hidden_units=[20, 20],
# The model must choose between 2 classes.
n_classes=2)
data_to_be_predicted, feature_id = get_features_to_be_predicted()
predictions = classifier.predict(
input_fn=lambda: predict_input_fn(data_to_be_predicted,
batch_size=args.batch_size)
)
predictions = list(predictions)
data_set = []
for i in range(0, len(predictions)):
features = data_to_be_predicted.values[i]
f_id = feature_id[i]
pre = predictions[i].get("class_ids")[0]
new_data_item = [f_id]
for feature_item in features:
new_data_item.append(feature_item)
new_data_item.append(pre)
data_set.append(new_data_item)
print(new_data_item)
print(data_set)
data_save.write_to_csv(
"data/y2.csv",
data=data_set,
header=None)
def pre_train():
print("[Pre-Train] Ready to do pre_train")
# Load training data from data/y1.csv
train_x, train_y = data_load.load_data_original_with_label(
"../data/y1.csv", columns.CSV_FEATURE_AND_LABEL_RESULT, "y1")
print("=====[View Train Data]=====")
print(train_x.head())
print("===========================")
# Get Feature Columns just from original_features.csv
# because original_features.csv only include features and feature_id
feature_columns = data_load.get_feature_column(
"../data/original_features.csv")
# Build 2 hidden layer DNN with 20, 20 units respectively.
checkpointing_config = tf.estimator.RunConfig(
save_checkpoints_secs=60, # Save checkpoints every 60 seconds.
keep_checkpoint_max=10, # Retain the 10 most recent checkpoints.
)
classifier = tf.estimator.DNNClassifier(
model_dir="../model/y_dnn",
config=checkpointing_config,
feature_columns=feature_columns,
# Two hidden layers of 20 nodes each.
hidden_units=[20, 20],
# The model must choose between 2 classes.
n_classes=2)
# Train the Model.
classifier.train(
input_fn=lambda: train_input_fn(train_x, train_y, batch_size=100),
steps=1000)
# Print where the model saved.
print("[Check Point]" + classifier.latest_checkpoint())