-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathptranking_wrapper.py
179 lines (149 loc) · 7.36 KB
/
ptranking_wrapper.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
import os
import pickle
import numpy as np
from ptranking.ltr_adhoc.eval.ltr import LTREvaluator
RESULT_FILE_NAME = 'result_summary.pkl'
class PtrankingWrapper:
def __init__(self, data_conf, weak_sup_conf, l2r_training_conf, result_path,
wl_kt_distance=None, individual_kt=None):
self.data_conf = data_conf
self.weak_sup_conf = weak_sup_conf
self.l2r_training_conf =l2r_training_conf
self.result_path = os.path.join(self.data_conf['project_root'], result_path)
self.debug = l2r_training_conf['debug']
self.ltr_evaluator = LTREvaluator()
self.wl_kt_distance = wl_kt_distance # save the kt distance between weak labels & true labels
self.individual_kt = individual_kt
''' using the default setting for loading dataset & using the default setting for evaluation '''
''' mainly parameters for ptranking package'''
self.ltr_evaluator.set_eval_setting(debug=self.debug, dir_output=result_path)
self.ltr_evaluator.set_data_setting(debug=self.debug, data_id=data_conf['dataset_name'],
dir_data=data_conf['processed_data_path'])
self.data_dict = self.ltr_evaluator.get_default_data_setting()
self.eval_dict = self.ltr_evaluator.get_default_eval_setting()
self.ltr_evaluator.set_scoring_function_setting(debug=self.debug, data_dict=self.data_dict)
self.ltr_evaluator.set_model_setting(debug=self.debug, model_id=l2r_training_conf['model']) # data_dict argument is required
self.model_para_dict = self.ltr_evaluator.get_default_model_setting()
self.sf_para_dict = self.ltr_evaluator.get_default_scoring_function_setting()
# model parameters setup in sf_para_dict
"""
Default
{'id': 'ffnns',
'ffnns': {'num_layers': 5, 'HD_AF': 'R', 'HN_AF': 'R', 'TL_AF': 'S', 'apply_tl_af': True, 'BN': True,
'RD': False, 'FBN': True, 'num_features': 10}}
```
"""
self.sf_para_dict['ffnns']['num_layers'] = 3
self.sf_para_dict['ffnns']['h_dim'] = 30
self.sf_para_dict['ffnns']['BN'] = True
self.sf_para_dict['ffnns']['FBN'] = True
self.sf_para_dict['ffnns']['apply_tl_af'] = False
self.data_dict['num_features'] = len(data_conf['features']) - 1 # -1: the label feature
self.data_dict['train_batch_size'] = l2r_training_conf['train_batch_size']
self.data_dict['test_batch_size'] = l2r_training_conf['test_batch_size']
self.eval_dict['epochs'] = l2r_training_conf['epochs']
self.ltr_evaluator.setup_eval(data_dict=self.data_dict, eval_dict=self.eval_dict,
sf_para_dict=self.sf_para_dict, model_para_dict=self.model_para_dict)
def set_data(self, X_train, X_test, Y_train, Y_test, qid_train=None, qid_test=None):
"""
Parameters
----------
X_train
X_test
Y_train
Y_test
Returns
-------
"""
print('Training data shape, X_train.shape', X_train.shape, 'Y_train.shape', Y_train.shape)
train_data, test_data, _ = self.ltr_evaluator.set_and_load_data(X_train=X_train, X_test=X_test,
Y_train=Y_train, Y_test=Y_test,
qid_train=qid_train,
qid_test=qid_test,
eval_dict=self.eval_dict,
data_dict=self.data_dict,
root_path=self.data_conf['project_root'])
self.train_data = train_data
self.test_data = test_data
def get_model(self):
"""
Get model based on parameter setup
Returns
-------
"""
model = self.ltr_evaluator.load_ranker(sf_para_dict=self.sf_para_dict, model_para_dict=self.model_para_dict,
opt=self.l2r_training_conf['optimizer'],
lr=self.l2r_training_conf['learning_rate'],
weight_decay=self.l2r_training_conf['weight_decay'])
return model
def train_model(self, model, IR=False, verbose=0, model_save=False):
"""
train model based on train_data, test_data
Parameters
----------
model
Returns
-------
"""
if IR:
ranker, result_summary = self.ltr_evaluator.custom_train_ir(ranker=model, eval_dict=self.eval_dict,
train_data=self.train_data,
test_data=self.test_data, verbose=verbose)
else:
ranker, result_summary = self.ltr_evaluator.custom_train(ranker=model, eval_dict=self.eval_dict,
train_data=self.train_data, test_data=self.test_data, verbose=verbose)
if model_save:
model_save_path = os.path.join(self.data_conf['project_root'], self.l2r_training_conf['model_checkpoint'], 'model.pkl')
ranker.save(dir=os.path.join(self.data_conf['project_root'], self.l2r_training_conf['model_checkpoint']),
name='model.pkl')
print('model saved in', model_save_path)
# save result
self.save_result(result_summary)
return result_summary
def eval(self, pred, verbose=0):
"""Evaluation with pred - it's for evaluation of label model itself
Args:
pred ([type]): Prediction, generated by weak labels
verbose (int, optional): [description]. Defaults to 0.
Returns:
[type]: [description]
"""
result_summary = self.ltr_evaluator.eval_with_pred(pred=pred, eval_dict=self.eval_dict,
train_data=self.train_data, test_data=self.test_data, verbose=verbose)
# save result
self.save_result(result_summary)
return result_summary
def load_model_checkpoint(self, model, save_path):
"""
load model checkpoint
Parameters
----------
model
save_path
Returns
-------
"""
model = model.load(save_path)
print("Model loaded from", save_path)
return model
def save_result(self, result_summary):
"""
Parameters
----------
result_summary
Returns
-------
"""
# append configurations to result_summary
result_summary['data_conf'] = self.data_conf
result_summary['weak_sup_conf'] = self.weak_sup_conf
result_summary['l2r_training_conf'] = self.l2r_training_conf
result_summary['wl_kt_distance'] = self.wl_kt_distance
result_summary['individual_kt'] = self.individual_kt
save_path = self.result_path
if not os.path.exists(save_path):
os.makedirs(save_path)
print("result data path", save_path, "generated")
with open(os.path.join(save_path, RESULT_FILE_NAME), 'wb') as fp:
pickle.dump(result_summary, fp)
print('The experiment result is saved in', os.path.join(save_path, 'result_summary.pkl'))