-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathvisualisation_evaluation_offline.py
208 lines (179 loc) · 11.3 KB
/
visualisation_evaluation_offline.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
import glob
import matplotlib.pyplot as plt
import numpy as np
import os
import torch
def plot(title, paths_results, ticks_x, label_x, show_legend=False, path_visual=None):
width_group = 0.8
size_group = 7
width_bar = width_group / size_group
x_group = np.linspace(-0.5 * width_group + width_bar, 0.5 * width_group, size_group)
colors = ["tab:blue", "tab:orange", "tab:green", "tab:red", "tab:purple", "tab:brown", "tab:pink", "tab:grey", "tab:olive", "tab:cyan"]
plt.rcParams.update({'font.size': 12})
plt.subplots_adjust(left=0.12, bottom=0.13, right=0.93)
for idx_path_result, path_result in enumerate(paths_results):
result = torch.load(path_result)
# true
plt.axhline(result["tsinf"]["return_eval_env_mean"], color="black", linestyle="--", label="true" if idx_path_result == 0 else "")
plt.axhline(np.NaN, color="none", label=" " if idx_path_result == 0 else "")
plt.axhline(np.NaN, color="none", label=" " if idx_path_result == 0 else "")
# tsinf
errorbar = plt.errorbar(idx_path_result + x_group[0], result["tsinf"]["return_eval_model_mean_mean"], yerr=2*(result["tsinf"]["return_eval_model_mean_std"] + result["tsinf"]["return_eval_model_std_mean"]), marker="s", markerfacecolor=colors[1], markeredgecolor="black", ecolor=colors[0], elinewidth=width_bar * 20, capsize=width_bar * 40)
errorbar[-1][0].set_linestyle("--")
errorbar = plt.errorbar(idx_path_result + x_group[0], result["tsinf"]["return_eval_model_mean_mean"], yerr=2*result["tsinf"]["return_eval_model_mean_std"], marker="s", markerfacecolor=colors[0], markeredgecolor="black", ecolor=colors[0], capsize=width_bar * 20)
errorbar[-1][0].set_linewidth(0)
plt.errorbar(idx_path_result + x_group[1], result["tsinf"]["return_eval_model_pessimistic_mean"], yerr=[[result["tsinf"]["return_eval_model_pessimistic_mean"] - result["tsinf"]["return_eval_model_pessimistic_min"]], [result["tsinf"]["return_eval_model_pessimistic_max"] - result["tsinf"]["return_eval_model_pessimistic_mean"]]], marker="D", markerfacecolor=colors[0], markeredgecolor="black", ecolor=colors[0], elinewidth=width_bar * 20, capsize=width_bar * 40)
errorbar[-1][0].set_linestyle("-")
# ts1
errorbar = plt.errorbar(idx_path_result + x_group[2], result["ts1_neutral"]["return_eval_model_mean"], yerr=2*result["ts1_neutral"]["return_eval_model_std"], marker="s", markerfacecolor=colors[1], markeredgecolor="black", ecolor=colors[1], elinewidth=width_bar * 20, capsize=width_bar * 40)
errorbar[-1][0].set_linestyle("--")
errorbar = plt.errorbar(idx_path_result + x_group[3], result["ts1_pessimistic"]["return_eval_model_mean"], yerr=[[result["ts1_pessimistic"]["return_eval_model_mean"] - result["ts1_pessimistic"]["return_eval_model_min"]], [result["ts1_pessimistic"]["return_eval_model_max"] - result["ts1_pessimistic"]["return_eval_model_mean"]]], marker="D", markerfacecolor=colors[1], markeredgecolor="black", ecolor=colors[1], elinewidth=width_bar * 20, capsize=width_bar * 40)
errorbar[-1][0].set_linestyle("-")
# ds
errorbar = plt.errorbar(idx_path_result + x_group[4], result["ds_neutral"]["return_eval_model_mean"], yerr=2*result["ds_neutral"]["return_eval_model_std"], marker="s", markerfacecolor=colors[2], markeredgecolor="black", ecolor=colors[2], elinewidth=width_bar * 20, capsize=width_bar * 40)
errorbar[-1][0].set_linestyle("--")
errorbar = plt.errorbar(idx_path_result + x_group[5], result["ds_pessimistic"]["return_eval_model_mean"], yerr=[[result["ds_pessimistic"]["return_eval_model_mean"] - result["ds_pessimistic"]["return_eval_model_min"]], [result["ds_pessimistic"]["return_eval_model_max"] - result["ds_pessimistic"]["return_eval_model_mean"]]], marker="D", markerfacecolor=colors[2], markeredgecolor="black", ecolor=colors[2], elinewidth=width_bar * 20, capsize=width_bar * 40)
errorbar[-1][0].set_linestyle("-")
plt.axvline(x=0.5, color="black", linewidth=0.1)
plt.axvline(x=1.5, color="black", linewidth=0.1)
plt.fill_between([-0.5,len(paths_results)-0.5], [2.0,2.0], [1.0,1.0], color="white", edgecolor="red", hatch="//")
plt.bar(np.NaN, np.NaN, color=colors[0], edgecolor="black", label="tsinf" if idx_path_result == 0 else "")
plt.bar(np.NaN, np.NaN, color=colors[1], edgecolor="black", label="ts1" if idx_path_result == 0 else "")
plt.bar(np.NaN, np.NaN, color=colors[2], edgecolor="black", label="ds" if idx_path_result == 0 else "")
plt.scatter(np.NaN, np.NaN, color="none", marker="s", edgecolor="black", label="neutral (baselines)" if idx_path_result == 0 else "")
plt.scatter(np.NaN, np.NaN, color="none", marker="D", edgecolor="black", label="conservative (ours)" if idx_path_result == 0 else "")
plt.title(title, fontsize=18)
plt.xticks(range(len(paths_results)), ticks_x)
plt.xlim(left=-0.5, right=len(paths_results)-0.5)
plt.xlabel(label_x, fontsize=18)
plt.yticks(np.arange(-0.5, 1.75, 0.5))
plt.ylim(bottom=-0.75, top=1.5)
plt.ylabel("Return", fontsize=15)
plt.gca().yaxis.set_label_coords(-0.08,0.5)
handles, labels = plt.gca().get_legend_handles_labels()
order = [0, 1, 2, 5, 6, 7, 3, 4]
if show_legend:
plt.legend([handles[idx] for idx in order], [labels[idx] for idx in order], ncol=3, fontsize=13)
if path_visual is None:
plt.show()
else:
plt.savefig(path_visual, format="pdf")
plt.clf()
if __name__ == "__main__":
dir_results = "Results"
dir_visuals = "Visuals"
label_x_data = "Dataset Size $|D_b|$"
label_x_horizon = "Horizon $T$"
configs = []
# pendulum data
config = {
"title": "Pendulum; Horizon $T = 200$",
"paths_results": ["results_pendulum_10000_noaleatoric_svgd10.0_25000_200", "results_pendulum_100000_noaleatoric_svgd10.0_25000_200", "results_pendulum_1000000_noaleatoric_svgd10.0_25000_200"],
"ticks_x": ["$10^4$", "$10^5$", "$10^6$"],
"label_x": label_x_data,
"show_legend": False,
"path_visual": os.path.join(dir_visuals, "visual_ope_pendulum_data_noaleatoric.pdf")
}
configs.append(config)
config = {
"title": "Pendulum; Horizon $T = 200$",
"paths_results": ["results_pendulum_10000_svgd10.0_25000_200", "results_pendulum_100000_svgd10.0_25000_200", "results_pendulum_1000000_svgd10.0_25000_200"],
"ticks_x": ["$10^4$", "$10^5$", "$10^6$"],
"label_x": label_x_data,
"show_legend": False,
"path_visual": os.path.join(dir_visuals, "visual_ope_pendulum_data.pdf")
}
configs.append(config)
# pendulum horizon
config = {
"title": "Pendulum; $|D_b| = 10^5$",
"paths_results": ["results_pendulum_100000_noaleatoric_svgd10.0_25000_100", "results_pendulum_100000_noaleatoric_svgd10.0_25000_200", "results_pendulum_100000_noaleatoric_svgd10.0_25000_400"],
"ticks_x": [100, 200, 400],
"label_x": label_x_horizon,
"show_legend": False,
"path_visual": os.path.join(dir_visuals, "visual_ope_pendulum_horizon_noaleatoric.pdf")
}
configs.append(config)
config = {
"title": "Pendulum; $|D_b| = 10^5$",
"paths_results": ["results_pendulum_100000_svgd10.0_25000_100", "results_pendulum_100000_svgd10.0_25000_200", "results_pendulum_100000_svgd10.0_25000_400"],
"ticks_x": [100, 200, 400],
"label_x": label_x_horizon,
"show_legend": False,
"path_visual": os.path.join(dir_visuals, "visual_ope_pendulum_horizon.pdf")
}
configs.append(config)
# hopper data
config = {
"title": "Hopper; Horizon $T = 200$",
"paths_results": ["results_hopper_10000_noaleatoric_svgd10.0_148000_200", "results_hopper_100000_noaleatoric_svgd10.0_148000_200", "results_hopper_1000000_noaleatoric_svgd10.0_148000_200"],
"ticks_x": ["$10^4$", "$10^5$", "$10^6$"],
"label_x": label_x_data,
"show_legend": True,
"path_visual": os.path.join(dir_visuals, "visual_ope_hopper_data_noaleatoric.pdf")
}
configs.append(config)
config = {
"title": "Hopper; Horizon $T = 200$",
"paths_results": ["results_hopper_10000_svgd10.0_148000_200", "results_hopper_100000_svgd10.0_148000_200", "results_hopper_1000000_svgd10.0_148000_200"],
"ticks_x": ["$10^4$", "$10^5$", "$10^6$"],
"label_x": label_x_data,
"show_legend": True,
"path_visual": os.path.join(dir_visuals, "visual_ope_hopper_data.pdf")
}
configs.append(config)
# hopper horizon
config = {
"title": "Hopper; $|D_b| = 10^5$",
"paths_results": ["results_hopper_100000_noaleatoric_svgd10.0_148000_100", "results_hopper_100000_noaleatoric_svgd10.0_148000_200", "results_hopper_100000_noaleatoric_svgd10.0_148000_400"],
"ticks_x": [100, 200, 400],
"label_x": label_x_horizon,
"show_legend": False,
"path_visual": os.path.join(dir_visuals, "visual_ope_hopper_horizon_noaleatoric.pdf")
}
configs.append(config)
config = {
"title": "Hopper; $|D_b| = 10^5$",
"paths_results": ["results_hopper_100000_svgd10.0_148000_100", "results_hopper_100000_svgd10.0_148000_200", "results_hopper_100000_svgd10.0_148000_400"],
"ticks_x": [100, 200, 400],
"label_x": label_x_horizon,
"show_legend": False,
"path_visual": os.path.join(dir_visuals, "visual_ope_hopper_horizon.pdf")
}
configs.append(config)
# halfcheetah data
config = {
"title": "HalfCheetah; Horizon $T = 100$",
"paths_results": ["results_halfcheetah_5_100000_noaleatoric_svgd10.0_wpm0.01_1000000_100", "results_halfcheetah_5_1000000_noaleatoric_svgd10.0_wpm0.01_1000000_100", "results_halfcheetah_5_4000000_noaleatoric_svgd10.0_wpm0.01_1000000_100"],
"ticks_x": ["$10^5$", "$10^6$", r"$4 \times 10^6$"],
"label_x": label_x_data,
"show_legend": False,
"path_visual": os.path.join(dir_visuals, "visual_ope_halfcheetah_data_noaleatoric.pdf")
}
configs.append(config)
# halfcheetah horizon
config = {
"title": r"Halfcheetah; $|D_b| = 4 \times 10^6$",
"paths_results": ["results_halfcheetah_5_4000000_noaleatoric_svgd10.0_wpm0.01_1000000_100", "results_halfcheetah_5_4000000_noaleatoric_svgd10.0_wpm0.01_1000000_150"],
"ticks_x": [100, 150],
"label_x": label_x_horizon,
"show_legend": False,
"path_visual": os.path.join(dir_visuals, "visual_ope_halfcheetah_horizon_noaleatoric.pdf")
}
configs.append(config)
# halfcheetah diversity
config = {
"title": r"Halfcheetah; Horizon $T = 100$; $|D_b| = 4 \times 10^6$",
"paths_results": ["results_halfcheetah_5_4000000_noaleatoric_svgd10.0_wpm0.01_1000000_100", "results_halfcheetah_12345_4000000_noaleatoric_svgd10.0_wpm0.01_1000000_100"],
"ticks_x": ["{5}", "{1,2,3,4,5}"],
"label_x": label_x_horizon,
"show_legend": True,
"path_visual": os.path.join(dir_visuals, "visual_ope_halfcheetah_diversity_noaleatoric.pdf")
}
configs.append(config)
# prepend results directory to lists of result paths
for idx_config in range(len(configs)):
for idx_path_result in range(len(configs[idx_config]["paths_results"])):
configs[idx_config]["paths_results"][idx_path_result] = os.path.join(dir_results, configs[idx_config]["paths_results"][idx_path_result])
for config in configs:
plot(**config)