-
Notifications
You must be signed in to change notification settings - Fork 2
/
generate_plots.py
349 lines (308 loc) · 11.7 KB
/
generate_plots.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
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
#!/usr/bin/env python
import os
import matplotlib as mpl
mpl.use('TkAgg')
import matplotlib.pyplot as plt
import matplotlib.axes as ax
import pandas as pd
import seaborn as sns
import argparse
import sys
try:
import cPickle as pickle
except ImportError:
import pickle
parser = argparse.ArgumentParser()
# general parameters
parser.add_argument('--env', type=str, required=True, help="which game to plot")
parser.add_argument('--max-timesteps', type=int, default=50)
parser.add_argument('--deep-rl-algo', type=str, default='a3c')
parser.add_argument('--result-folder', type=str, default='a3c')
parser.add_argument('--saveplot', action='store_true', help='save plot or not')
parser.set_defaults(saveplot=False)
parser.add_argument('--nolegend', action='store_true', help='dont show legend')
parser.set_defaults(nolegend=False)
parser.add_argument('--folder', type=str, default='plots', help="where to save the plot")
parser.add_argument('--fn', type=str, default='', help="plot name")
# A3CTBSIL and LiDER
parser.add_argument('--baseline', action='store_true', help='plot A3CTBSIL')
parser.set_defaults(baseline=False)
parser.add_argument('--lidera3c', action='store_true', help='plot LiDER')
parser.set_defaults(lidera3c=False)
# Ablation studies
parser.add_argument('--sampler', action='store_true', help='LiDER-SampleR')
parser.set_defaults(sampler=False)
parser.add_argument('--onebuffer', action='store_true', help='LiDER-OneBuffer')
parser.set_defaults(onebuffer=False)
parser.add_argument('--addall', action='store_true', help='LiDER-AddAll')
parser.set_defaults(addall=False)
# Extentions
parser.add_argument('--liderta', action='store_true', help='LiDER-TA')
parser.set_defaults(liderta=False)
parser.add_argument('--liderbc', action='store_true', help='LiDER-BC')
parser.set_defaults(liderbc=False)
args = parser.parse_args()
game = args.env
deep_rl_algo = args.deep_rl_algo
game_env_type = 'NoFrameskip_v4'
location = 'lower right'
ncol = 1
num_data = [1, 2, 3] # detault three trials
# plot the horizontal line for TA and BC
# NOTE: their values need to be set manually
# when running LiDER-TA or LiDER-BC, the pretrained models are evaluated for 50 episodes at step 0
# their results are saved under pretrained_models/TA(orBC)/[game]/[game]-modeleval.txt
plot_ta = True if args.liderta else False
plot_bc = True if args.liderbc else False
# set legend location (if needed)
if game == 'Freeway':
pass
# location = "upper left"
elif game == 'Gopher':
location = "upper left"
elif game == 'MsPacman':
# pass
location = "upper left"
elif game == 'NameThisGame':
# pass
location = "upper left"
elif game =='Alien':
location = "upper left"
elif game == 'MontezumaRevenge':
location = "upper left"
else:
print("Invalid game!!!")
sys.exit()
gym_env = game + game_env_type
print(gym_env)
#sns.set_style("darkgrid")
sns.set(context='paper', style='darkgrid', rc={'figure.figsize':(7,5)})
LW = 1.5
ALPHA = 0.1
MARKERSIZE = 12
plt.figure(figsize=(6.6,4.8))
N = 1
if deep_rl_algo == 'a3c':
MAX_TIMESTEPS = args.max_timesteps*1000000
else:
print("Invalid algorithm!!!")
sys.exit()
PER_EPOCH = 1000000
def create_dataframe(rewards_all_trials, per_epoch=0, max_timesteps=0, label=''):
if per_epoch == 0:
per_epoch = PER_EPOCH
if max_timesteps == 0:
max_timesteps = MAX_TIMESTEPS
timestep = [ t/per_epoch for t in range(0, (max_timesteps+per_epoch), per_epoch) ]
d = {}
print("steps," + ','.join(map(str, sorted(rewards_all_trials[0]['eval'].keys()))))
for data_idx, reward_data in enumerate(rewards_all_trials):
rewards = []
all_rewards = []
for r in sorted(reward_data['eval'].keys()):
all_rewards.append(reward_data['eval'][r][0])
if r <= max_timesteps and r % per_epoch == 0:
rewards.append(reward_data['eval'][r][0]) ### <=== (reward, steps, num_episodes)
print(label + '-' + str(data_idx+1) + '-' + str(len(all_rewards)) + ':' + ','.join(map(str, all_rewards)) + '\n')
d['Rewards{}'.format(data_idx+1)] = rewards
df = pd.DataFrame(data=d, index=timestep)
df.index.name = 'Timestep'
df = df.iloc[::N]
return df
def plot_fun(
ex_type='', color='green', result_folder='a3c',
marker='o', markersize=MARKERSIZE,
lw=LW, linestyle=None, label='test', num_data=[1,2,3]):
''' creates dataframe and plots graph '''
rewards_all_trials = []
for data_idx in num_data:
folder=('results/{}/'.format(result_folder) + gym_env + \
'{}_{}/'.format(ex_type, data_idx) + \
gym_env + '-{}-rewards.pkl'.format(deep_rl_algo))
print(folder)
r_data = pickle.load(open(folder, 'rb'))
rewards_all_trials.append(r_data)
df_rewards = create_dataframe(rewards_all_trials, label=label)
while len(rewards_all_trials):
del rewards_all_trials[0]
df_rewards_mean = df_rewards.mean(axis=1)
print ("MEAN: ", df_rewards.mean(axis=0))
df_rewards_std = df_rewards.std(axis=1)
plt.plot(
df_rewards_mean.index,
df_rewards_mean,
color=sns.xkcd_rgb[color],
marker=marker,
markersize=markersize,
markevery=3,
lw=lw,
linestyle=linestyle,
label=label)
plt.fill_between(
df_rewards_std.index,
df_rewards_mean - df_rewards_std,
df_rewards_mean + df_rewards_std,
color=sns.xkcd_rgb[color],
alpha=ALPHA)
def plot_ta_bc_fun(timestep, mean_val, std_val, color='purple', lw=2,
linestyle='dotted', alpha=0.1, label=None):
plt.plot(
TIMESTEP,
[mean_val]*len(TIMESTEP),
color=sns.xkcd_rgb[color],
lw=2,
linestyle=linestyle,
label=label)
plt.fill_between(
TIMESTEP,
mean_val - std_val,
mean_val + std_val,
color=sns.xkcd_rgb[color],
alpha=0.08)
################plotting start here##########################
# A3CTBSIL
if args.baseline:
ex_type = '_rawreward_transformedbell_sil_prioritymem'
plot_fun(
ex_type=ex_type,
color='violet red',
marker='P',
result_folder=args.result_folder,
label='{}TBSIL'.format(deep_rl_algo.upper()),
num_data=num_data)
ncol += 1
# LiDER
if args.lidera3c:
ex_type = '_rawreward_transformedbell_sil_prioritymem_lider'
plot_fun(
ex_type=ex_type,
color='green',#'kelly green',
marker='^',#'x',
result_folder=args.result_folder,
label='LiDER',
num_data=num_data)
ncol += 1
# LiDER-AddAll
if args.addall:
ex_type = '_rawreward_transformedbell_sil_prioritymem_lider_addall'
plot_fun(
ex_type=ex_type,
color='dark sky blue',
marker='d',
result_folder=args.result_folder,
label='LiDER-AddAll',
num_data=num_data)
ncol += 1
# LiDER-OneBuffer
if args.onebuffer:
ex_type = '_rawreward_transformedbell_sil_prioritymem_lider_onebuffer'
plot_fun(
ex_type=ex_type,
color='forest green',
marker='.',
result_folder=args.result_folder,
label='LiDER-OneBuffer',
num_data=num_data)
ncol += 1
# LiDER-SampleR
if args.sampler:
ex_type = '_rawreward_transformedbell_sil_prioritymem_lider_sampleR'
plot_fun(
ex_type=ex_type,
color='purple',
marker='x',
result_folder=args.result_folder,
label='LiDER-SampleR',
num_data=num_data)
ncol += 1
# LiDER-TA
if args.liderta:
ex_type='_rawreward_transformedbell_sil_prioritymem_lider_TA'
plot_fun(
ex_type=ex_type,
color='orange',
marker='o',
result_folder=args.result_folder,
label='LiDER-TA',
num_data=num_data)
ncol += 1
# LiDER-BC
if args.liderbc:
ex_type = '_rawreward_transformedbell_sil_prioritymem_lider_BC'
plot_fun(
ex_type=ex_type,
color='cobalt',
marker='v',
result_folder=args.result_folder,
label='LiDER-BC',
num_data=num_data)
ncol += 1
TIMESTEP = [ t/PER_EPOCH for t in range(0, (MAX_TIMESTEPS+PER_EPOCH), PER_EPOCH) ]
# horizontal lines
# values taken from the paper; replace with your experiment results
if plot_ta:
ncol += 1
if args.env == 'MsPacman':
plot_ta_bc_fun(TIMESTEP, mean_val=9145.42, std_val=955.94, color='purple', lw=2, linestyle='dotted')
plt.text(x=30, y=10500, s='Trained Agent', fontsize='xx-large',color='purple')
elif args.env == 'Alien':
plot_ta_bc_fun(TIMESTEP, mean_val=7190.4, std_val=1251.27, color='purple', lw=2, linestyle='dotted')
# plt.text(x=30, y=7100, s='Trained Agent', fontsize='xx-large',color='purple')
elif args.env == 'MontezumaRevenge':
plot_ta_bc_fun(TIMESTEP, mean_val=1108.0, std_val=1057.14, color='purple', lw=2, linestyle='dotted')
# plt.text(x=32, y=1400, s='Trained Agent', fontsize='xx-large',color='purple')
elif args.env == 'Freeway':
plot_ta_bc_fun(TIMESTEP, mean_val=32.92, std_val=0.27, color='purple', lw=2, linestyle='dotted')
# plt.text(x=30, y=33.45, s='Trained Agent', fontsize='xx-large',color='purple')
elif args.env == 'NameThisGame':
plot_ta_bc_fun(TIMESTEP, mean_val=9969.0, std_val=1910.91, color='purple', lw=2, linestyle='dotted')
# plt.text(x=30, y=10000, s='Trained Agent', fontsize='xx-large',color='purple')
elif args.env == 'Gopher':
plot_ta_bc_fun(TIMESTEP, mean_val=6972.4, std_val=2190.26, color='purple', lw=2, linestyle='dotted')
# plt.text(x=34, y=7100, s='Trained Agent', fontsize='xx-large',color='purple')
else:
print("Invalid game!!!")
sys.exit()
if plot_bc:
ncol += 1
if args.env == 'MsPacman':
plot_ta_bc_fun(TIMESTEP, mean_val=1776.6, std_val=993.94, color='black', lw=2, linestyle='--')
plt.text(x=30, y=1100, s='Behavior Cloning', fontsize='xx-large',color='black')
elif args.env == 'Alien':
plot_ta_bc_fun(TIMESTEP, mean_val=839.2, std_val=718.72, color='black', lw=2, linestyle='--')
# plt.text(x=30, y=900, s='Behavior Cloning', fontsize='xx-large',color='black')
elif args.env == 'MontezumaRevenge':
plot_ta_bc_fun(TIMESTEP, mean_val=174.0, std_val=205.73, color='black', lw=2, linestyle='--')
# plt.text(x=32, y=250, s='Behavior Cloning', fontsize='xx-large',color='black')
elif args.env == 'Freeway':
plot_ta_bc_fun(TIMESTEP, mean_val=25.06, std_val=1.48, color='black', lw=2, linestyle='--')
# plt.text(x=30, y=25.5, s='Behavior Cloning', fontsize='xx-large',color='black')
elif args.env == 'NameThisGame':
plot_ta_bc_fun(TIMESTEP, mean_val=1491.2, std_val=530.55, color='black', lw=2, linestyle='--')
# plt.text(x=30, y=1500, s='Behavior Cloning', fontsize='xx-large',color='purple')
elif args.env == 'Gopher':
plot_ta_bc_fun(TIMESTEP, mean_val=450.8, std_val=393.27, color='black', lw=2, linestyle='--')
# plt.text(x=30, y=700, s='Behavior Cloning', fontsize='xx-large',color='black')
else:
print("Invalid game!!!")
sys.exit()
plt.tick_params(axis='y',labelsize='x-large')
plt.xlabel('Steps (in millions)', fontsize='x-large')
plt.ylabel('Reward', fontsize='x-large')
if args.env=="NameThisGame": # make sure all plots start from 0
plt.gca().set_ylim(bottom=0)
if not args.nolegend:
plt.legend(loc=location, fontsize='xx-large', ncol=1)
header = game
plt.title('{}'.format(game), fontsize='xx-large')
if args.saveplot:
assert args.fn != '', "must provide file names to save the plot"
figname = '{}_{}_'.format(game, deep_rl_algo)
figname += args.fn
path = args.folder + '/' + args.env
if not os.path.exists(path):
os.makedirs(path)
plt.tight_layout()
plt.savefig('{}/{}/{}.pdf'.format(args.folder, args.env, figname))
plt.savefig('{}/{}/{}.png'.format(args.folder, args.env, figname))
plt.show()