-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
177 lines (119 loc) · 5.62 KB
/
utils.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
import os
import time
from datetime import timedelta
import matplotlib.pyplot as plt
import matplotlib.ticker as plticker
import loss
import tensorflow as tf
import numpy as np
class History:
"""This class represents the training history of a model. It can load the
prior history when training continues, keeps track of the training and
validation error, and finally plots them as a curve after each epoch.
"""
def __init__(self, n_train_batches, n_valid_batches,
dataset, path, device):
self.train_history = []
self.valid_history = []
self._prior_epochs = 0
self._train_error = 0
self._valid_error = 0
self._n_train_batches = n_train_batches
self._n_valid_batches = n_valid_batches
self._path = path
self._id = (dataset, device)
self._get_prior_history()
def _get_prior_history(self):
if os.path.isfile(self._path + "train_%s_%s.txt" % self._id):
with open(self._path + "train_%s_%s.txt" % self._id, "r") as file:
for line in file.readlines():
self.train_history.append(float(line))
if os.path.isfile(self._path + "valid_%s_%s.txt" % self._id):
with open(self._path + "valid_%s_%s.txt" % self._id, "r") as file:
for line in file.readlines():
self.valid_history.append(float(line))
self.prior_epochs = len(self.train_history)
def update_train_step(self, train_error):
self._train_error += train_error
def update_valid_step(self, valid_error):
self._valid_error += valid_error
def get_mean_train_error(self, reset=True):
mean_train_error = self._train_error / self._n_train_batches
if reset:
self._train_error = 0
return mean_train_error
def get_mean_valid_error(self, reset=True):
mean_valid_error = self._valid_error / self._n_valid_batches
if reset:
self._valid_error = 0
return mean_valid_error
def save_history(self):
mean_train_loss = self.get_mean_train_error(False)
mean_valid_loss = self.get_mean_valid_error(False)
self.train_history.append(mean_train_loss)
self.valid_history.append(mean_valid_loss)
os.makedirs(self._path, exist_ok=True)
with open(self._path + "train_%s_%s.txt" % self._id, "a") as file:
file.write("%f\n" % self.train_history[-1])
with open(self._path + "valid_%s_%s.txt" % self._id, "a") as file:
file.write("%f\n" % self.valid_history[-1])
if len(self.train_history) > 1:
axes = plt.figure().gca()
x_range = np.arange(1, len(self.train_history) + 1)
plt.plot(x_range, self.train_history, label="train", linewidth=2)
plt.plot(x_range, self.valid_history, label="valid", linewidth=2)
plt.legend()
plt.xlabel("epochs")
plt.ylabel("error")
locations = plticker.MultipleLocator(base=1.0)
axes.xaxis.set_major_locator(locations)
plt.savefig(self._path + "curve_%s_%s.png" % self._id)
plt.close()
class Progbar:
"""This class represents a progress bar for the terminal that visualizes
the training progress for each epoch, estimated time of accomplishment,
and then summarizes the training and validation loss together with the
elapsed time.
"""
def __init__(self, n_train_data, n_train_batches,
batch_size, n_epochs, prior_epochs):
self._train_time = 0
self._valid_time = 0
self._start_time = time.time()
self._batch_size = batch_size
self._n_train_data = n_train_data
self._n_train_batches = n_train_batches
self._target_epoch = str(n_epochs + prior_epochs).zfill(2)
self._current_epoch = str(prior_epochs + 1).zfill(2)
def _flush(self):
self._train_time = 0
self._valid_time = 0
self._start_time = time.time()
current_epoch_int = int(self._current_epoch) + 1
self._current_epoch = str(current_epoch_int).zfill(2)
def update_train_step(self, current_batch):
current_batch += 1
self._train_time = time.time() - self._start_time
batch_train_time = self._train_time / current_batch
eta = (self._n_train_batches - current_batch) * batch_train_time
eta = str(timedelta(seconds=np.ceil(eta)))
progress_line = "=" * (20 * current_batch // self._n_train_batches)
current_instance = current_batch * self._batch_size
current_instance = np.clip(current_instance, 0, self._n_train_data)
progress_frac = "%i/%i" % (current_instance, self._n_train_data)
information = (self._current_epoch, self._target_epoch,
progress_line, progress_frac, eta)
progbar_output = "Epoch %s/%s [%-20s] %s (ETA: %s)" % information
print(progbar_output, end="\r", flush=True)
def update_valid_step(self):
self._valid_time = time.time() - self._start_time - self._train_time
def write_summary(self, mean_train_loss, mean_valid_loss):
train_time = str(timedelta(seconds=np.ceil(self._train_time)))
valid_time = str(timedelta(seconds=np.ceil(self._valid_time)))
train_information = (mean_train_loss, train_time)
valid_information = (mean_valid_loss, valid_time)
train_output = "\n\tTrain loss: %.6f (%s)" % train_information
valid_output = "\tValid loss: %.6f (%s)" % valid_information
print(train_output, flush=True)
print(valid_output, flush=True)
self._flush()