-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathplotting.py
186 lines (146 loc) · 7 KB
/
plotting.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
import sys
import math
import matplotlib
matplotlib.use("Agg")
import numpy as np
from scipy.interpolate import UnivariateSpline
import matplotlib.pyplot as plt
import pandas as pd
from collections import OrderedDict
class Plotter:
def __init__(self):
self.values = OrderedDict()
self.losses = DataResolutionManager()
def add_values(self, values):
"""
Accepts a list of tuples (score_name, score) and adds it to the internal representation to be plotted
:param scores: a list of tuples (score_name, score)
:return: None
"""
for val in values:
try:
self.values[val[0]].append(val[1])
except KeyError:
self.values[val[0]] = DataResolutionManager()
self.values[val[0]].append(val[1])
def add_loss(self, losses):
self.losses.append(losses)
def plot(self, title):
fig, axes = plt.subplots(nrows=2, ncols=1)
if self.values:
lines = []
max_length = max([len(values[1].get_values()) for values in self.values.items()])
for key in self.values:
values = self.__scale__(self.values[key].get_values(), max_length+1)
series = pd.Series(values, name=key)
lines.append(series)
df = pd.DataFrame(lines)
# DEPRECATED
else:
line1_values = self.losses.get_values()
line2_values = self.scores.get_values()
line2_values = self.__scale__(line2_values, len(line1_values))
line3_values = self.second_scores.get_values()
line3_values = self.__scale__(line3_values, len(line1_values))
if len(line1_values) == 0 and len(line2_values) == 0:
raise Exception("Cannot plot empty values losses and scores")
line1 = pd.Series(line1_values, name=self.line1_name)
line2 = pd.Series(line2_values, name=self.line2_name)
line3 = pd.Series(line3_values, name=self.line3_name)
df = pd.DataFrame([line1, line2, line3])
if len(self.losses) > 0:
line = pd.Series(self.losses.get_values(), name="Loss")
loss_frame = pd.DataFrame(line)
loss_ax = loss_frame.plot(ax=axes[0], title=title, legend=True, figsize=(16, 9))
loss_ax.xaxis.set_ticks([i * self.num_episodes / 10 for i in range(11)])
loss_ax.set_xticklabels([i * self.num_episodes / 10 for i in range(11)])
else:
plt.title(title)
df = df.transpose()
ax = df.plot(ax=axes[1], legend=True, figsize=(16, 9), ylim=(-1, 1) if len(self.losses) > 0 else None) # secondary_y=[line2_name] for separate scales | ylim=(min, max) for limiting y scale
ax.xaxis.set_ticks([i * self.num_episodes / 10 for i in range(11)])
ax.set_xticklabels([i * self.num_episodes / 10 for i in range(11)])
ax.yaxis.set_ticks = [i/2 for i in range(-2, 3)]
ax.axhline(y=0, color='black', linewidth=2)
ax.axhline(y=0.5, color='black', linewidth=1)
ax.axhline(y=-0.5, color='black', linewidth=1)
plt.xlabel("Episodes")
plt.grid()
return plt
@staticmethod
def __scale__(lst, length):
if len(lst) > 1:
old_indices = np.arange(0, len(lst))
new_indices = np.linspace(0, len(lst) - 1, length)
spl = UnivariateSpline(old_indices, lst, k=1, s=0)
lst = spl(new_indices)
return lst
@property
def num_episodes(self):
return max((len(self.values[value]) for value in self.values))
class DataResolutionManager:
def __init__(self, data_points=[], storage_size=1000):
try: # data_points can be either DataResolutionManagers or simple lists
data_points = data_points.get_values()
except AttributeError:
pass
data_points = data_points.copy()
self.storage_size = storage_size
self.compression_factor = math.ceil((len(data_points)+1) / storage_size)
self.values = []
self.buffer = []
if self.compression_factor > 1:
self.buffer = data_points[len(data_points) - len(data_points) % self.compression_factor:]
for i in range(len(data_points) // self.compression_factor):
self.values.append(sum(data_points[:self.compression_factor]) / self.compression_factor)
data_points = data_points[self.compression_factor:]
else:
self.values = data_points
def append(self, value):
self.buffer.append(value)
if len(self.buffer) >= self.compression_factor:
self.values.append(sum(self.buffer) / len(self.buffer))
self.buffer = []
if len(self.values) >= 2*self.storage_size:
if len(self.values) % 2 != 0: # Move uneven element back to buffer
self.buffer.append(self.values.pop())
self.values = [(a + b) / 2 for a, b in zip(self.values[0::2], self.values[1::2])]
self.compression_factor *= 2
def get_values(self):
if len(self.buffer) == 0:
return self.values
else:
return self.values + [sum(self.buffer) / len(self.buffer)]
def get_real_length(self):
return len(self.values) * self.compression_factor + len(self.buffer)
def __str__(self):
return self.get_values().__str__()
def __len__(self):
return len(self.values)
class Printer:
@staticmethod
def print_episode(episode, total_episodes, time_taken=None, print_every_iteration=False):
"""
Prints progress every full percent of the episode.
:param episode: The current episode number
:param total_episodes: The total episode number
:param time_taken: A datetime diff eg. start_time - datetime.now()
:param print_every_iteration: Flag that lets progress be printed every episode instead of only on full percentages
:return: True if progress was printed, False otherwise
"""
""" Prints progress on the current episode.
Only prints full percentages unless specified otherwise using the @print_every_iteration flag"""
if print_every_iteration or 100 * episode/total_episodes % 1 == 0:
Printer.print_inplace("Episode %s/%s" % (episode, total_episodes), 100 * episode // total_episodes, time_taken)
return True
return False
@staticmethod
def print_inplace(text, percentage, time_taken=None, comment=""):
percentage = int(percentage)
length_factor = 5
progress_bar = int(round(percentage/length_factor)) * "*" + (round((100-percentage)/length_factor)) * "."
progress_bar = progress_bar[:round(len(progress_bar)/2)] + "|" + str(int(percentage)) + "%|" + progress_bar[round(len(progress_bar)/2):]
sys.stdout.write("\r%s |%s|" % (text, progress_bar) + (" Time: %s" % str(time_taken).split(".")[0] if time_taken else "") + comment)
sys.stdout.flush()
if percentage == 100:
print()