forked from SMARTlab-Purdue/PrefMMT
-
Notifications
You must be signed in to change notification settings - Fork 0
/
visualize.py
207 lines (166 loc) · 6.71 KB
/
visualize.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
import pandas as pd
import numpy as np
import os
import glob
import matplotlib
import matplotlib.pyplot as plt
font = 'Arial'
plt.rcParams['figure.dpi'] = 300
plt.rcParams['font.family'] = font
plt.rcParams['mathtext.fontset'] = 'custom'
plt.rcParams['mathtext.rm'] = font
plt.rcParams['mathtext.it'] = font
plt.rcParams['mathtext.bf'] = font
plt.rcParams['axes.linewidth'] = 0.5
plt.rcParams['xtick.major.width'] = 0.5
plt.rcParams['xtick.minor.width'] = 0.5
plt.rcParams['ytick.major.width'] = 0.5
plt.rcParams['ytick.minor.width'] = 0.5
linewidth = 2.5
# import tensorflow as tf
import tensorboard as tb
from tensorboard.backend.event_processing import event_accumulator
print("TensorBoard version: ", tb.__version__)
PINK = (247/255, 112/255, 136/255)
GREEN = (51/255, 176/255, 122/255)
BLUE = (128/255, 150/255, 244/255)
BLUEBLUE = (0, 83/255, 214/255)
YELLOW = (255/255, 161/255, 0/255)
BLACK = (0, 0, 0)
# https://yeun.github.io/open-color/#red
VIOLET9 = (95/255, 61/255, 196/255)
PINK9 = (166/255, 30/255, 77/255)
GRAY9 = (33/255, 37/255, 41/255)
GRAY8 = (52/255, 58/255, 64/255)
GRAY7 = (73/255, 80/255, 87/255)
ORANGE7 = (247/255, 103/255, 7/255)
GRAY6 = (134/255, 142/255, 150/255)
GRAY4 = (206/255, 212/255, 218/255)
RED4 = (255/255, 135/255, 135/255)
PINK4 = (247/255, 131/255, 172/255)
GRAPE4 = (218/255, 119/255, 242/255)
VIOLET4 = (151/255, 117/255, 250/255)
INDIGO4 = (116/255, 143/255, 252/255)
BLUE4 = (77/255, 171/255, 247/255)
CYAN4 = (59/255, 201/255, 219/255)
TEAL4 = (56/255, 217/255, 169/255)
GREAN4 = (105/255, 219/255, 124/255)
LIME4 = (169/255, 227/255, 75/255)
YELLOW4 = (255/255, 212/255, 59/255)
ORANGE4 = (255/255, 169/255, 77/255)
# COLOR_LIST = [GRAY7, GRAPE4, VIOLET4, BLUE4, TEAL4, LIME4, YELLOW4, ORANGE4, RED4]
COLOR_LIST = [RED4, ORANGE4, YELLOW4, LIME4, TEAL4, INDIGO4, VIOLET4, GRAPE4, GRAY7, PINK, GREEN, BLUE, YELLOW, BLACK]
# COLOR_LIST = [GRAY7, VIOLET4, RED4, TEAL4, YELLOW4, GRAPE4, LIME4, BLUE4, ORANGE4]
def load_df_from_tb_event(tb_event, col='evaluation/average_returns'):
ea = event_accumulator.EventAccumulator(tb_event)
ea.Reload()
try:
df = pd.DataFrame(ea.Scalars(col))
except:
print(f"tb_event: {tb_event}")
raise
return df[['step', 'value']]
def get_data_from_all_seeds(tb_file_list, col='evaluation/avearge_returns', window=1):
df = None
for tb_file in tb_file_list:
if df is None:
# Dirty and quick fix to incorporate
# for csv data from KH (eval every 10000)
# and tensorboard log from JS (eval every 40000).
try:
df = pd.read_csv(tb_file)
df = df.rename(columns={'Step': 'step', 'Value': 'value'})
df = df[['step', 'value']]
df = df[df.index % window == 0]
except:
df = load_df_from_tb_event(tb_file, col=col)
else:
try:
append_df = pd.read_csv(tb_file)
append_df = append_df.rename(columns={'Step': 'step', 'Value': 'value'})
append_df = append_df[['step', 'value']]
df = pd.concat([df, append_df], axis=1)
df = df[df.index % window == 0]
except:
df = pd.concat([df, load_df_from_tb_event(tb_file, col=col)], axis=1)
return df
def exp_smooth(df, alpha=0.4):
return df['value'].ewm(alpha=alpha).mean()
def rolling(df, window=4):
return df['value'].rolling(window, min_periods=1).mean()
def mean_std(df):
df_mean = df.mean(axis=1)
df_std = df.std(axis=1)
return df_mean, df_mean - df_std, df_mean + df_std
def process_data(tb_list, col='evaluation/average_returns', verbose=True, window=1):
df_list = get_data_from_all_seeds(tb_list, col=col, window=window)
if verbose:
print(df_list)
smoothed_mean, smoothed_under_std, smoothed_over_std = mean_std(rolling(df_list, window=window))
x = df_list['step'].iloc[:, 1].to_numpy()
y_mean = smoothed_mean.to_numpy()
y_under_std = smoothed_under_std.to_numpy()
y_over_std = smoothed_over_std.to_numpy()
return x, y_mean, y_under_std, y_over_std
def draw_graph(title='',
xlim_lower=0,
xlim_upper=1000000,
ylim_upper=100,
ylim_lower=0,
fill_density=0.15,
figsize=(5, 3.5),
idx=201,
verbose=False,
no_legend=False,
save=True,
save_path='./graphs/',
show_title=True,
show_var=True,
legend_loc='upper left',
color_list=COLOR_LIST,
col='evaluation/average_returns',
extension='png',
**kwargs,
):
line_num = 0
label_list = []
xticks = np.linspace(xlim_lower, xlim_upper, 5)
yticks = np.linspace(ylim_lower, ylim_upper, 5)
for key, value in kwargs.items():
if 'label' in key:
label_list.append(value)
fill_density = fill_density
_, ax = plt.subplots(1, 1, figsize=figsize, dpi=500)
for key, value in kwargs.items():
if 'tb_list' in key:
xx, yy_mean, yy_under_std, yy_over_std = process_data(value, col=col, verbose=verbose)
ax.plot(xx[:idx], yy_mean[:idx], color=color_list[line_num], label=label_list[line_num], linewidth=linewidth * 1.25)
if show_var:
ax.fill_between(xx[:idx], yy_under_std[:idx], yy_over_std[:idx], facecolor=(*color_list[line_num], fill_density), edgecolor=(0, 0, 0, 0))
print(f"{label_list[line_num]}: {yy_mean[-1]:.4f} ± {yy_mean[-1] - yy_under_std[-1]:.4f}")
line_num += 1
ax.set_xlabel('Training Step', fontsize=14)
ax.set_ylabel('Average Return', fontsize=14)
if show_title:
ax.set_title(title, fontsize=16)
ax.grid(alpha=1.0, linestyle=':', linewidth=0.25)
ax.tick_params(axis='both', which='major', labelsize=12)
ax.set_yticks(yticks)
ax.set_xticks(xticks)
ax.set_xticks([100000, 300000, 500000, 700000, 900000], minor=True)
def set_xtick(x, p):
return '{}$\\times 10^5$'.format(int(x / 100000))
# NOTE: use xtick with 10^4 or xlabel with 10^4
ax.get_xaxis().set_major_formatter(
matplotlib.ticker.FuncFormatter(set_xtick)
)
ax.xaxis.major.formatter._useMathText = True
ax.set_xlim(xlim_lower, xlim_upper)
ax.set_ylim(ylim_lower, ylim_upper)
if not no_legend:
leg = ax.legend(fancybox=False, fontsize=8, edgecolor='black', borderaxespad=0.1, handlelength=1.5, loc=legend_loc)
leg.get_frame().set_linewidth(0.5)
plt.tight_layout()
if save:
os.makedirs(save_path, exist_ok=True)
plt.savefig(save_path + '/' + title + f".{extension}")