-
Notifications
You must be signed in to change notification settings - Fork 10
/
dl_util.py
214 lines (181 loc) · 7.85 KB
/
dl_util.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
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
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
import os
import warnings
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
from os.path import join
def warn(*args, **kwargs):
pass
warnings.filterwarnings("ignore")
warnings.warn = warn
import sys
import platform
import datetime
import h5py
import time
import numpy as np
if platform.system() == 'Linux':
sys.path.append("/home/apx748/OPV")
else:#It is Mac/Darwin
sys.path.append("/Users/arindam/OPV")
if platform.python_version().split(".")[0] == "2":
from keras.layers import Merge
import tensorflow as tf
import keras
from keras.models import Sequential
from keras.layers import Dense, Dropout, LSTM, GRU, Bidirectional, TimeDistributed, Flatten
from keras.layers.embeddings import Embedding
from keras.layers.convolutional import Conv1D, MaxPooling1D
from keras.wrappers.scikit_learn import KerasRegressor
from keras.models import model_from_json
from keras.preprocessing import sequence
from keras.preprocessing.text import Tokenizer
import pickle
import matplotlib.pyplot as plt
from ml_util import str_round, loadData
def as_keras_metric(method):
import functools
from keras import backend as K
import tensorflow as tf
@functools.wraps(method)
def wrapper(self, args, **kwargs):
""" Wrapper for turning tensorflow metrics into keras metrics """
value, update_op = method(self, args, **kwargs)
K.get_session().run(tf.local_variables_initializer())
with tf.control_dependencies([update_op]):
value = tf.identity(value)
return value
return wrapper
precision = as_keras_metric(tf.metrics.precision)
recall = as_keras_metric(tf.metrics.recall)
auc = as_keras_metric(tf.metrics.auc)
def save_history(history,history_file,path):
obj = {"history":history.history, "epoch":history.epoch}
with open(path+"/"+history_file + '.pkl', 'wb') as f:
pickle.dump(obj, f)
print("History saved successfully", history_file+".pkl")
def save_model(model_type, model, history, dropout, epochs=10, batch_size=32, lr=0.001, prefix=None):
if prefix is None:
file_suffix = "_"+model_type+"_dropout_"+str(dropout)\
+"_epochs_"+str(epochs)+"_batch_"+str(batch_size)+"_lr_"+str(lr)
else:
file_suffix = "_"+prefix+"_"+model_type+"_dropout_"+str(dropout)\
+"_epochs_"+str(epochs)+"_batch_"+str(batch_size)+"_lr_"+str(lr)
model_file = join("model","model"+file_suffix+".json")
weights_file = join("model","weights"+file_suffix+".h5")
history_file = "history"+file_suffix
model.save_weights(weights_file)
# serialize model to JSON
model_json = model.to_json()
with open(model_file, "w") as json_file:
json_file.write(model_json)
save_history(history,history_file,"model")
print("Model saved successfully in", model_file)
print("Weights saved successfully in", weights_file)
def prepare_message(model_type, stats, dropout, epochs, batch_size, lr, time_elapsed, size=1, ml_task = "regression", prefix=None):
message = "Type of model: "
message += model_type+"\n\n"
if ml_task == "classification":
message += "AUC score: "+str_round(stats["auc"])+"\n"
message += "Precision: "+str_round(stats["precision"])+"\n"
message += "Recall: "+str_round(stats["recall"])+"\n"
message += "Accuracy: "+str_round(stats["accuracy"])+"\n\n"
else:
#stats = {"mape":mape, "mae":mean_absolute_err, "mse":mean_squared_err, "r2":r2}
message += "Mean absolute percent error: "+str_round(stats["mape"])+" %\n"
message += "Mean absolute error: "+str_round(stats["mae"])+"\n"
message += "Mean squared error: "+str_round(stats["mse"])+"\n"
message += "R^2 : "+str_round(stats["r2"])+"\n\n"
if size==1:
message+="Whole dataset was used\n\n"
else:
message+=str(size*100)+" % of the dataset was used\n\n"
message+="Training took "+str(time_elapsed)+" seconds\n\n"
if prefix is None:
file_suffix = "_"+model_type+"_dropout_"+str(dropout)\
+"_epochs_"+str(epochs)+"_batch_"+str(batch_size)+"_lr"+str(lr)
else:
file_suffix = "_"+prefix+"_"+model_type+"_dropout_"+str(dropout)\
+"_epochs_"+str(epochs)+"_batch_"+str(batch_size)+"_lr"+str(lr)
model_file = join("model","model"+file_suffix+".json")
weights_file = join("model","weights"+file_suffix+".h5")
history_file = join("model","history"+file_suffix)
stats_file = join("model","stats"+file_suffix)
message+="Model saved successfully in "+ model_file+"\n"
message+="Weights saved successfully in "+ weights_file+"\n"
message+="History saved successfully in "+ history_file+".pkl\n"
message+="Stats saved successfully in "+ stats_file+".pkl\n"
return message
def generate_onehot_encoding(docs,type="encoded_docs"):
t = Tokenizer(char_level=True)
t.fit_on_texts(docs)
if type=="vocab_size":
vocab_size = len(t.word_index) + 1
return vocab_size
encoded_docs = t.texts_to_sequences(docs)#, mode='count')
return encoded_docs
def plot_history(history, metric="mean_absolute_percentage_error"):
plt.figure()
plt.xlabel('Epoch')
metric_label = ""
for item in metric.split("_"):
metric_label = metric_label + " "+ item.capitalize()
if "absolute" in metric:
plt.ylabel(metric_label+" %")
else:
plt.ylabel(metric_label)
if "error" in metric.lower():
label_train, label_val = "Train Error", "Validation Error"
elif "auc" in metric.lower():
label_train, label_val = "Train AUC", "Validation AUC"
elif "precision" in metric.lower():
label_train, label_val = "Train Precision", "Validation Precision"
elif "recall" in metric.lower():
label_train, label_val = "Train Recall", "Validation Recall"
else:
label_train, label_val = "Train Accuracy", "Validation Accuracy"
plt.plot(history.epoch, np.array(history.history[metric]),
label=label_train)
plt.plot(history.epoch, np.array(history.history['val_'+metric]),
label =label_val)
plt.legend()
if "auc" in metric or "acc" in metric or "precision" in metric or "recall" in metric:
max_lim = 1
else:
max_lim = max(max(np.array(history.history[metric])),\
max(np.array(history.history['val_'+metric])))*2
if max_lim>100:
max_lim = 100
plt.ylim([0, max_lim])
def plot_history_from_file(history_file, metric="mean_absolute_percentage_error"):
history = loadData(history_file,"model")
plt.figure()
plt.xlabel('Epoch')
metric_label = ""
for item in metric.split("_"):
metric_label = metric_label + " "+ item.capitalize()
if "absolute" in metric:
plt.ylabel(metric_label+" %")
else:
plt.ylabel(metric_label)
if "error" in metric.lower():
label_train, label_val = "Train Error", "Validation Error"
elif "auc" in metric.lower():
label_train, label_val = "Train AUC", "Validation AUC"
elif "precision" in metric.lower():
label_train, label_val = "Train Precision", "Validation Precision"
elif "recall" in metric.lower():
label_train, label_val = "Train Recall", "Validation Recall"
else:
label_train, label_val = "Train Accuracy", "Validation Accuracy"
plt.plot(history.epoch, np.array(history["history"][metric]),
label='Train Error')
plt.plot(history.epoch, np.array(history["history"]['val_'+metric]),
label = 'Val Error')
plt.legend()
if "auc" in metric or "r2" in metric.lower() or "acc" in metric or "precision" in metric or "recall" in metric:
max_lim = 1
else:
max_lim = max(max(np.array(history["history"][metric])),\
max(np.array(history["history"]['val_'+metric])))*2
if max_lim>100:
max_lim = 100
plt.ylim([0, max_lim])