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
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
|
- import warnings
- import logging
- import pandas as pd
- import numpy as np
- from argparse import ArgumentParser
- from tensorflow.keras.models import Sequential
- from tensorflow.keras.layers import Dense
- from tensorflow.keras.optimizers import Adam
- from tensorflow.keras.layers import LeakyReLU
- from sklearn.preprocessing import StandardScaler
- from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score
- from sklearn.model_selection import train_test_split
- import mlflow.sklearn
- import mlflow.keras
- logging.basicConfig(level=logging.DEBUG)
- logger = logging.getLogger(__name__)
- def load_data(path):
- try:
- df = pd.read_csv(path, decimal=",", parse_dates=["date"], infer_datetime_format=True)
- return df
- except Exception as exp:
- logger.exception("Unable to load the data. Error: %s", exp)
- def data_split(df):
- try:
- x_tr, x_te, y_train, y_test = train_test_split(df, df['% Silica Concentrate'],
- test_size=0.2,
- random_state=42)
- x_tr.drop(labels=['% Silica Feed', "Flotation Column 03 Air Flow", 'date', '% Silica Concentrate'],
- axis=1,
- inplace=True)
- x_te.drop(labels=['% Silica Feed', "Flotation Column 03 Air Flow", 'date', '% Silica Concentrate'],
- axis=1,
- inplace=True)
- return x_tr, x_te, y_train, y_test
- except Exception as exp:
- logger.exception("Error: %s", exp)
- def data_preprocessing(train_set, test_set):
- try:
- scaler = StandardScaler()
- x_train_arr = scaler.fit_transform(train_set)
- x_test_arr = scaler.fit_transform(test_set)
- x_train_df = pd.DataFrame(x_train_arr, columns=train_set.columns)
- x_test_df = pd.DataFrame(x_test_arr, columns=test_set.columns)
- return x_train_df, x_test_df
- except Exception as exp:
- logger.exception("Error: %s", exp)
- def model_training(x_train_df, y_train, alph, lr, epoch, batch):
- try:
- model = Sequential()
- model.add(Dense(50, input_dim=20, kernel_initializer='normal'))
- model.add(LeakyReLU(alpha=alph))
- model.add(Dense(50, kernel_initializer='normal'))
- model.add(LeakyReLU(alpha=alph))
- model.add(Dense(50, kernel_initializer='normal'))
- model.add(LeakyReLU(alpha=alph))
- model.add(Dense(50, kernel_initializer='normal'))
- model.add(LeakyReLU(alpha=alph))
- model.add(Dense(50, kernel_initializer='normal'))
- model.add(LeakyReLU(alpha=alph))
- model.add(Dense(50, kernel_initializer='normal'))
- model.add(LeakyReLU(alpha=alph))
- model.add(Dense(1, kernel_initializer='normal'))
- # Compile model
- opt = Adam(lr=lr)
- model.compile(loss='mean_squared_error', optimizer=opt, metrics=['mse'])
- mlflow.keras.autolog()
- # Fit the model
- history = model.fit(x_train_df, y_train, validation_split=0.2, epochs=epoch, batch_size=batch, verbose=2)
- return model, history
- except Exception as exp:
- logger.exception("Error: %s", exp)
- def eval_metrics(actual, predictions):
- try:
- msq_err = mean_squared_error(actual, predictions)
- mean_abs_err = mean_absolute_error(actual, predictions)
- regress_coeff = r2_score(actual, predictions)
- return msq_err, mean_abs_err, regress_coeff
- except Exception as exp:
- logger.exception("Error: %s", exp)
- if __name__ == "__main__":
- warnings.filterwarnings("ignore")
- np.random.seed(40)
- parser = ArgumentParser(description="Production Optimization Example")
- parser.add_argument(
- "--max_epochs", type=int, default=2, help="Training Iterations"
- )
- parser.add_argument("--batch_size", type=int, default=20, help="Batch size")
- parser.add_argument(
- "--learning_rate", type=float, default=0.0003, help="Learning rate"
- )
- parser.add_argument(
- "--alpha", type=float, default=0.1, help="Alpha rate"
- )
- args = parser.parse_args()
- dict_args = vars(args)
- data_frame = load_data("mining_flotation.csv")
- logger.info("Data is loaded Successfully!")
- train, test, y_tr, y_te = data_split(data_frame)
- logger.info("Data is split into training and testing sets")
- x_train, x_test = data_preprocessing(train, test)
- logger.info("Normalized the training and testing sets")
- alpha = dict_args["alpha"]
- lr = dict_args["learning_rate"]
- epochs = dict_args["max_epochs"]
- batch = dict_args["batch_size"]
- with mlflow.start_run(run_name=f'Model Training: Epochs = {epochs}'):
- model_r, result = model_training(x_train, y_tr, alpha, lr, epochs, batch)
- logger.info("Model has been successfully trained and tracked with MLFlow")
- mlflow.log_param("alpha", alpha)
- mlflow.log_param("Learning Rate", lr)
- mlflow.log_param("Epochs", epochs)
- mlflow.log_param("Batch Size", batch)
- y_pred = model_r.predict(x_test)
- logger.info("Model Predictions")
- coeff_1, coeff_2, coeff_3 = eval_metrics(y_te, y_pred)
- logger.info("Evaluation metrics are computed and published")
- mlflow.log_metric("mse", coeff_1)
- mlflow.log_metric("r2", coeff_3)
- mlflow.log_metric("mae", coeff_2)
- mlflow.keras.log_model(model_r, "models")
- mlflow.end_run()
|