-
Notifications
You must be signed in to change notification settings - Fork 36
/
Copy pathsample_stat_summary.py
150 lines (131 loc) · 5.87 KB
/
sample_stat_summary.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
#coding=utf-8
import sys
import json
import numpy as np
file_name = "output_data.json"
output_file_names = ["correct_times", "correct_ratio", "avg_probs", "label_var",
"max_label_probs", "min_label_probs", "forgetting_times", "learnt_times", "first_forget", "first_learn", "pred_label", "pred_dist"]
num_samples = 100
_input_path = "./outputs/"
_output = open(_input_path + file_name + ".result", "w")
def list_concat(score_dict, input_file, input_path = "./data/", sample_size=1, pred_idx=-2, label_idx=-1, score_idx=-3, get_max_probs=False):
"""
add info of each epoch (or given steps) into a dict of lists
"""
_input = open(input_path + input_file, "r")
for i, line in enumerate(_input):
info = json.loads(line.strip().replace("\'", "\""))
sid = int(info["id"])
label = int(info["label"]) # [1:-1] to avoid "[]"
if "noisy_label" in info:
s_label = int(info["noisy_label"])
else:
s_label = 0
# score = float(info["probs"])
label_probs = float(info["label_probs"]) # the score under GL
pred_correctness = info["correct"]
if get_max_probs:
all_probs = [eval(j) for j in info["probs"]]
max_probs = np.max(all_probs)
else:
max_probs = 1.0
score_info = [] # list of scores under different class
for score in info["probs"]: # the number of classes here
score_float = float(score)
score_info.append(score_float)
if not score_dict["id"][sid]:
score_dict["id"][sid] = sid
score_dict["label"][sid] = label
score_dict["s_label"][sid] = s_label
score_dict["label_probs"][sid].append(label_probs)
score_dict["max_probs"][sid].append(max_probs)
score_dict["pred_info"][sid].append(pred_correctness)
score_dict["pred_label"][sid].append(str(np.argmax(score_info)))
# add forget info
list_length = len(score_dict["pred_info"][sid])
if list_length > 1:
if score_dict["pred_info"][sid][list_length - 1] == score_dict["pred_info"][sid][list_length - 2]:
score_dict["forget_info"][sid].append("None")
elif score_dict["pred_info"][sid][list_length - 1] == "true":
score_dict["forget_info"][sid].append("Learn")
else:
score_dict["forget_info"][sid].append("Forget")
else:
score_dict["forget_info"][sid].append("None")
#if sid == 1:
# print(score_dict["forget_info"][sid])
# if i >= sample_size:
# break
_input.close()
def check_correct_ratio(correct_lists):
"""
ratio that a model predict classes correctly in different epochs
"""
if len(correct_lists) == 0 or len(correct_lists[0]) == 0:
return [0], [0]
ratio_list = []
pos_list = []
for c_list in correct_lists:
pos_cnt = 0
for info in c_list:
if info == "true":
pos_cnt += 1
ratio_list.append(float(pos_cnt)/len(c_list) if len(c_list)!=0 else 0)
pos_list.append(pos_cnt)
return pos_list, ratio_list
def check_forget_time(forget_lists):
if len(forget_lists) == 0 or len(forget_lists[0]) == 0:
return [0], [0], 0, 0
forgetting_list = []
learnt_list = []
first_forgetting_time = []
first_learnt_time = []
for f_list in forget_lists:
forgetting_cnt = 0
learnt_cnt = 0
first_f_time = 0
first_l_time = 0
for i, info in enumerate(f_list):
if info == "Forget":
forgetting_cnt += 1
if first_f_time == 0:
first_f_time = i
elif info == "Learn":
learnt_cnt += 1
if first_l_time == 0:
first_l_time = i
forgetting_list.append(forgetting_cnt)
learnt_list.append(learnt_cnt)
first_forgetting_time.append(first_f_time)
first_learnt_time.append(first_l_time)
return forgetting_list, learnt_list, first_forgetting_time, first_learnt_time
def check_pred_distribution(pred_lists):
pred_list = []
for scores in pred_lists:
score_dist_dict = {"0": 0, "1": 0, "2": 0, "3": 0, "4": 0}
for score in scores:
score_dist_dict[score] += 1
pred_list.append(score_dist_dict)
return pred_list
info_dict = {"id": [[] for i in range(num_samples)], "label": [[] for i in range(num_samples)],
"s_label": [[] for i in range(num_samples)], "label_probs": [[] for i in range(num_samples)],
"max_probs": [[] for i in range(num_samples)], "pred_info": [[] for i in range(num_samples)],
"forget_info": [[] for i in range(num_samples)], "pred_label": [[] for i in range(num_samples)]}
list_concat(info_dict, file_name, _input_path, sample_size=num_samples)
print(len(info_dict["label_probs"]), len(info_dict["label_probs"][0]))
info_dict["correct_times"], info_dict["correct_ratio"] = check_correct_ratio(info_dict["pred_info"])
info_dict["label_var"] = np.var(info_dict["label_probs"], axis=1)
info_dict["max_var"] = np.var(info_dict["max_probs"], axis=1)
info_dict["avg_probs"] = np.mean(info_dict["label_probs"], axis=1)
info_dict["max_label_probs"] = np.max(info_dict["label_probs"], axis=1)
info_dict["min_label_probs"] = np.min(info_dict["label_probs"], axis=1)
info_dict["forgetting_times"], info_dict["learnt_times"], info_dict["first_forget"], info_dict["first_learn"] = check_forget_time(info_dict["forget_info"])
info_dict["pred_dist"] = check_pred_distribution(info_dict["pred_label"])
output_file_names = ["id", "label", "s_label"] + output_file_names
_output.write("\t".join(output_file_names) + "\n")
for i in range(num_samples):
info_list = []
for name in output_file_names:
info_list.append(str(info_dict[name][i]))
_output.write("\t".join(info_list) + "\n")
_output.close()