-
Notifications
You must be signed in to change notification settings - Fork 1
/
validate_submission.py
213 lines (163 loc) · 7.19 KB
/
validate_submission.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
import argparse
import os
from os.path import join as pjoin
import pandas as pd
from evaluation import Evaluator
import zipfile
import shutil
debug_mode = 0
def mkdir(d):
if not os.path.exists(d):
os.makedirs(d)
def build_ref_pred_pair(ref_dict, pred_dict):
ref_list, pred_list = [], []
for k, v in ref_dict.items():
ref_list.append([v])
if k in pred_dict:
pred_list.append(pred_dict[k])
else:
pred_list.append(' ')
return ref_list, pred_list
def get_evaluations_final(run_mf, qrel):
metrics = {'recall_10', 'ndcg_cut_10'}
eval_obj = Evaluator(metrics)
indiv_res = eval_obj.evaluate(run_mf, qrel)
overall_res = eval_obj.show_all()
return overall_res, indiv_res
def read_run_file(run_file, score_columns='score'):
qret = {}
df_qret = pd.read_csv(run_file, sep="\t")
for row in df_qret.itertuples():
cur_user_qret = qret.get(str(row.userId), {})
cur_user_qret[str(row.itemId)] = float(row.score)
qret[str(row.userId)] = cur_user_qret
return qret
def read_qrel_file(qrel_file):
qrel = {}
df_qrel = pd.read_csv(qrel_file, sep="\t")
for row in df_qrel.itertuples():
cur_user_qrel = qrel.get(str(row.userId), {})
cur_user_qrel[str(row.itemId)] = int(row.rating)
qrel[str(row.userId)] = cur_user_qrel
return qrel
def read_string(solution_file):
with open(solution_file) as fi:
return fi.read().strip()
def merge_run_files(run_dir, market1, market2, output_market):
predict_path_val_market1 = os.path.join(run_dir, market1, 'valid_pred.tsv')
predict_path_test_market1 = os.path.join(run_dir, market1, 'test_pred.tsv')
predict_path_val_market2 = os.path.join(run_dir, market2, 'valid_pred.tsv')
predict_path_test_market2 = os.path.join(run_dir, market2, 'test_pred.tsv')
output_market_dir_path = os.path.join(run_dir, output_market)
mkdir(output_market_dir_path)
predict_path_val_out = os.path.join(run_dir, output_market, 'valid_pred.tsv')
predict_path_test_out = os.path.join(run_dir, output_market, 'test_pred.tsv')
write_market_files(predict_path_val_market1, predict_path_val_market2, predict_path_val_out)
write_market_files(predict_path_test_market1, predict_path_test_market2, predict_path_test_out)
def write_market_files(predict_path_val_market1, predict_path_val_market2, predict_path_val_out):
with open(predict_path_val_market1) as fi1:
with open(predict_path_val_market2) as fi2:
with open(predict_path_val_out, 'w') as fo:
for l in fi1:
fo.write(l)
for l in fi2:
if not l.startswith('userId'):
fo.write(l)
def validate_file_structure(extract_dir):
for m in ['t1', 't2']:
for f in ['test_pred.tsv', 'valid_pred.tsv']:
try:
with open(os.path.join(extract_dir, m, f)) as fi:
pass
except FileNotFoundError:
print('{} not found!'.format(os.path.join(extract_dir, m, f)))
return False
return True
def get_scores_for_market(input_dir, data_dir, market_name):
# prepare for val set
predict_path_val = os.path.join(input_dir, market_name, 'valid_pred.tsv')
ref_path_val = os.path.join(data_dir, market_name, 'valid_qrel.tsv')
my_valid_run = read_run_file(predict_path_val)
my_valid_qrel = read_qrel_file(ref_path_val)
task_ov_val, task_ind_val = get_evaluations_final(my_valid_run, my_valid_qrel)
return task_ov_val
def offline_scores(df, score_columns, gt_path, market_name):
'''
df: valid_df.
score_columns: columns need to evaluation.
gt_path: path of groudtruth.
print:
nDCG@10 and recall@10
'''
new_df = df.copy()
new_df['score'] = df[score_columns]
get_df_scores_for_market(new_df, gt_path, market_name)
def parse_predict_df(df_qret):
qret = {}
for row in df_qret.itertuples():
cur_user_qret = qret.get(str(row.userId), {})
cur_user_qret[str(row.itemId)] = float(row.score)
qret[str(row.userId)] = cur_user_qret
return qret
def get_df_scores_for_market(predict_df, data_dir, market_name):
# prepare for val set
ref_path_val = os.path.join(data_dir, 'valid_qrel.tsv')
my_valid_run = parse_predict_df(predict_df)
my_valid_qrel = read_qrel_file(ref_path_val)
task_ov_val, task_ind_val = get_evaluations_final(my_valid_run, my_valid_qrel)
scores = ['ndcg_cut_10', 'recall_10']
score_names = {
'recall_10': {'val': 'r10_val', 'test': 'r10_test'},
'ndcg_cut_10': {'val': 'ndcg10_val', 'test': 'ndcg10_test'}
}
print(
"===================== Market : " + market_name + "=====================")
for score in scores: # iterating over the scores
score_val_name = score_names[score]['val']
score_val = task_ov_val[score]
print(
"======= Set val : score(" + score_val_name + ")=%0.12f =======" % score_val)
return task_ov_val
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--submission_file", help="Zip file that contains the run files to be submitted to Codalab.")
parser.add_argument("--data_dir", help="Path to the DATA dir of the kit. Default: ./DATA/.", default='./DATA/')
args = parser.parse_args()
extract_dir = './tmp/'
scores = ['ndcg_cut_10', 'recall_10']
score_names = {
'recall_10': {'val': 'r10_val', 'test': 'r10_test'},
'ndcg_cut_10': {'val': 'ndcg10_val', 'test': 'ndcg10_test'}
}
# We assume that the submission comes with two markets (i.e., t1 and t2).
marekts = ['t1', 't2', 't1t2']
# First we unzip the run file in a tmp folder then start evaluating it.
mkdir(extract_dir)
print('Extracting the submission zip file')
with zipfile.ZipFile(args.submission_file, "r") as zip_ref:
zip_ref.extractall(extract_dir)
print('Validating the file structure of the submission')
file_structure_validation = validate_file_structure(extract_dir)
if file_structure_validation:
print('File structure validation successfully passed')
else:
print('File structure validation failed. Please refer to the instructions')
return
print('Evaluating the validation set')
# Then we merge the run files of the two markets for the joint performance evaluation, and call it 't1t2'.
merge_run_files(extract_dir, 't1', 't2', 't1t2')
task_ov_test, task_ov_val = {}, {}
for m in marekts: # iterate over the three target markets (including the joint one)
print(
"===================== Market : " + m + "=====================")
task_ov_val[m] = get_scores_for_market(extract_dir, args.data_dir, m)
for score in scores: # iterating over the scores
score_val_name = score_names[score]['val']
score_val = task_ov_val[m][score]
print(
"======= Set val : score(" + score_val_name + ")=%0.12f =======" % score_val)
# remove the tmp directory
shutil.rmtree(extract_dir)
# python validate_submission.py --submission_file ../result/tmptime_12162307/submission.zip --data_dir ../../input
if __name__ == "__main__":
main()