2021-05-12 09:37:42 +02:00
|
|
|
import sys
|
2021-05-13 10:48:52 +02:00
|
|
|
|
2021-05-12 09:37:42 +02:00
|
|
|
import mlflow
|
|
|
|
import pandas as pd
|
|
|
|
from sklearn.metrics import mean_squared_error
|
|
|
|
from sklearn.model_selection import train_test_split
|
|
|
|
from tensorflow.keras.callbacks import EarlyStopping
|
|
|
|
from tensorflow.keras.layers import Dense, Dropout
|
|
|
|
from tensorflow.keras.models import Sequential
|
2021-05-23 17:12:13 +02:00
|
|
|
from urllib.parse import urlparse
|
|
|
|
|
|
|
|
mlflow.set_experiment("s430705")
|
|
|
|
mlflow.set_tracking_uri("http://172.17.0.1:5000")
|
2021-05-12 09:37:42 +02:00
|
|
|
|
|
|
|
|
2021-05-12 09:42:24 +02:00
|
|
|
def prepare_model(train_size_param, test_size_param, epochs, batch_size):
|
2021-05-12 09:37:42 +02:00
|
|
|
movies_data = pd.read_csv("train.csv", error_bad_lines=False)
|
|
|
|
movies_data.drop(movies_data.columns[0], axis=1, inplace=True)
|
|
|
|
movies_data.dropna(inplace=True)
|
|
|
|
X = movies_data.drop("rating", axis=1)
|
|
|
|
Y = movies_data["rating"]
|
|
|
|
|
|
|
|
X_train, X_test, Y_train, Y_test = train_test_split(
|
|
|
|
X, Y, test_size=test_size_param, random_state=42
|
|
|
|
)
|
|
|
|
|
|
|
|
test_df = pd.read_csv("test.csv")
|
|
|
|
test_df.drop(test_df.columns[0], axis=1, inplace=True)
|
|
|
|
x_test = test_df.drop("rating", axis=1)
|
|
|
|
y_test = test_df["rating"]
|
|
|
|
|
|
|
|
# Set up model
|
|
|
|
model = Sequential()
|
|
|
|
model.add(Dense(8, activation="relu"))
|
|
|
|
model.add(Dropout(0.5))
|
|
|
|
model.add(Dense(3, activation="relu"))
|
|
|
|
model.add(Dropout(0.5))
|
|
|
|
model.add(Dense(1))
|
|
|
|
model.compile(optimizer="adam", loss="mse")
|
|
|
|
|
|
|
|
early_stop = EarlyStopping(monitor="val_loss", mode="min", verbose=1, patience=10)
|
|
|
|
|
|
|
|
model.fit(
|
|
|
|
x=X_train.values,
|
|
|
|
y=Y_train.values,
|
|
|
|
validation_data=(X_test, Y_test.values),
|
|
|
|
batch_size=batch_size,
|
|
|
|
epochs=epochs,
|
|
|
|
callbacks=[early_stop],
|
|
|
|
)
|
|
|
|
y_pred = model.predict(x_test.values)
|
|
|
|
|
2021-05-23 12:33:10 +02:00
|
|
|
input_example = X_test.values[10]
|
|
|
|
|
2021-05-12 09:37:42 +02:00
|
|
|
rmse = mean_squared_error(y_test, y_pred)
|
|
|
|
|
|
|
|
model.save("model_movies")
|
|
|
|
|
2021-05-23 12:33:10 +02:00
|
|
|
return model, rmse, X_train, input_example
|
2021-05-12 09:37:42 +02:00
|
|
|
|
|
|
|
|
|
|
|
train_size_param = float(sys.argv[1]) if len(sys.argv) > 1 else 0.8
|
|
|
|
test_size_param = float(sys.argv[2]) if len(sys.argv) > 1 else 0.2
|
|
|
|
epochs = int(sys.argv[3]) if len(sys.argv) > 1 else 400
|
|
|
|
batch_size = int(sys.argv[4]) if len(sys.argv) > 1 else 128
|
|
|
|
|
|
|
|
|
|
|
|
with mlflow.start_run():
|
|
|
|
|
|
|
|
mlflow.log_param("train size", train_size_param)
|
|
|
|
mlflow.log_param("test size", test_size_param)
|
|
|
|
mlflow.log_param("epochs", epochs)
|
|
|
|
mlflow.log_param("batch size", batch_size)
|
|
|
|
|
2021-05-23 12:33:10 +02:00
|
|
|
model, rmse, X_train, input_example = prepare_model(
|
2021-05-12 09:37:42 +02:00
|
|
|
train_size_param=train_size_param,
|
|
|
|
test_size_param=test_size_param,
|
|
|
|
epochs=epochs,
|
|
|
|
batch_size=batch_size,
|
|
|
|
)
|
|
|
|
|
|
|
|
mlflow.log_metric("RMSE", rmse)
|
|
|
|
|
2021-05-23 12:33:10 +02:00
|
|
|
signature = mlflow.models.signature.infer_signature(X_train.values, model.predict(X_train.values))
|
|
|
|
mlflow.keras.save_model(model, "movies_imdb", input_example=input_example, signature=signature)
|
2021-05-23 17:12:13 +02:00
|
|
|
|
|
|
|
tracking_url_type_store = urlparse(mlflow.get_tracking_uri()).scheme
|
|
|
|
if tracking_url_type_store != "file":
|
|
|
|
mlflow.keras.log_model(model, "movies_imdb", registered_model_name="s430705", signature=signature,
|
|
|
|
input_example=input_example)
|
|
|
|
else:
|
|
|
|
mlflow.keras.log_model(model, "model_movies", signature=signature, input_example=input_example)
|
|
|
|
mlflow.keras.save_model(model, "movies_mdb", signature=signature, input_example=input_example)
|