-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrun_tests2.0.py
153 lines (135 loc) · 6.41 KB
/
run_tests2.0.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
from collections import defaultdict
from sklearn.metrics.pairwise import cosine_distances
import numpy as np
import pandas as pd
import paths
from utils.openai_helpers import get_embedding_caller
from modules.sim_dataset import SimDataset
from modules.fasttext_wrapper import FasttextWrapper
from modules.supabase_objects import get_one_contract_example_data
from modules.sim_predictor import SimPredictor
def custom_sim_metric(df):
"""
returns count of when
1. identical pair has larger distance than similar of different pair
2. similar pair has larger distance than different pair
"""
count = 0
for i, row in df.iterrows():
label, distance = row['label'], row['Distance']
if label == 'I':
count += len(df[df['label'].isin(['S', 'D']) & (df['Distance'] < distance)])
elif row['label'] == 'S':
count += len(df[df['label'].isin(['D']) & (df['Distance'] < distance)])
return count
def compose_csv(label_to_vector_pairs_info, output_name):
rows = []
for label, vector_pairs_info in label_to_vector_pairs_info.items():
for vector_pair_info in vector_pairs_info:
rows.append({
'Function A': vector_pair_info['a_f'],
'Function B': vector_pair_info['b_f'],
'Description of difference': vector_pair_info['descr'],
'Distance': vector_pair_info['distance'],
'label': label,
})
rows.sort(key=lambda x: x['Distance'])
df = pd.DataFrame(rows)
score = custom_sim_metric(df)
csv_path = f"{paths.RESULTS_DIR}/{output_name}_score_{score}.csv"
df.to_csv(csv_path)
print(f"saved {csv_path}")
def prepare_test_data():
csv_path = f"{paths.DATA_DIR}/testset_updated.csv"
df = pd.read_csv(csv_path)
a_functions, b_functions, labels, descriptions = [], [], [], []
for index, row in df.iterrows():
a_functions.append(row.to_dict()["Original Function"])
b_functions.append(row.to_dict()["Modified Function"])
labels.append(row.to_dict()["(I)dentical/(S)imilar/(D)ifferent"])
descriptions.append(row.to_dict()["Explanation of changes"])
a_data = {"functions": [{"function_str": f, "encoded_ir": ""} for f in a_functions]}
b_data = {"functions": [{"function_str": f, "encoded_ir": ""} for f in b_functions]}
return {
'a_data': a_data,
'b_data': b_data,
'a_functions': a_functions,
'b_functions': b_functions,
'labels': labels,
'descriptions': descriptions
}
def get_label_to_details(data, a_vectors, b_vectors, distance_function):
label_to_vector_pairs_info = defaultdict(list)
for av, bv, a_f, b_f, label, descr in zip(
a_vectors, b_vectors, data['a_functions'],
data['b_functions'], data['labels'], data['descriptions']):
distance = distance_function(av.reshape(1, -1), bv.reshape(1, -1))[0][0]
label_to_vector_pairs_info[label].append({
'a_f': a_f,
'b_f': b_f,
'descr': descr,
'distance': round(distance, 4),
})
return label_to_vector_pairs_info
def func_test_fasttext(configs, output_filename='sim_test_results_fasttext'):
df = pd.read_csv(f"{paths.DATA_DIR}/testset_updated.csv")
originals = [{'function_str': row['Original Function']} for i, row in df.iterrows()]
modifieds = [{'function_str': row['Modified Function']} for i, row in df.iterrows()]
test_data = prepare_test_data()
data = SimDataset.load(configs.get('data_path'))
wrapper = FasttextWrapper.load(data, configs.get('full_model_name'))
distance_function = configs.get('distance_function')
a_vectors, b_vectors = [], []
for original, modified in zip(originals, modifieds):
a_vectors.append(wrapper.get_sentence_vector(original))
b_vectors.append(wrapper.get_sentence_vector(modified))
label_to_vector_pairs_info = get_label_to_details(test_data, a_vectors, b_vectors, distance_function)
compose_csv(label_to_vector_pairs_info, output_filename)
def func_test_openai(configs, output_filename='sim_test_results_openai'):
get_embedding = get_embedding_caller()
df = pd.read_csv(f"{paths.DATA_DIR}/testset_updated.csv")
originals = [row['Original Function'] for i, row in df.iterrows()]
modifieds = [row['Modified Function'] for i, row in df.iterrows()]
test_data = prepare_test_data()
distance_function = configs.get('distance_function')
a_vectors, b_vectors = [], []
for original, modified in zip(originals, modifieds):
a_vectors.append(np.array(get_embedding(original)))
b_vectors.append(np.array(get_embedding(modified)))
label_to_vector_pairs_info = get_label_to_details(test_data, a_vectors, b_vectors, distance_function)
compose_csv(label_to_vector_pairs_info, output_filename)
def contract_test_fasttext(configs):
ex_data = get_one_contract_example_data()
data = SimDataset.load(configs.get('data_path'))
predictor = SimPredictor({
'models': {
'function_str': 'fasttext',
# 'encoded_ir': 'fasttext',
},
'dataset_name': configs.get('dataset_name')
})
contr_id_to_sim, func_strs = predictor.get_contr_id_to_sim(ex_data)
id_contr = data['id_contracts']
rows = []
for (contract_id, score), _ in zip(sorted(
contr_id_to_sim.items(), key=lambda x: x[1], reverse=False), range(10)):
contr = id_contr[contract_id]
rows.append([contr['id'], round(score, 3), contr['contract_str'],
func_strs, ex_data['contract_str']])
pd.DataFrame(rows, columns=['contract_id', 'score', 'contract_str',
'input_functions', 'example_contract_str'])\
.to_csv(f'{paths.RESULTS_DIR}/contract_sim.csv')
if __name__ == '__main__':
###########################################################################
###########################################################################
test_configs = {
'full_model_name': 'fasttext_function_str',
'dataset_name': 'processed_data',
'data_path': f"{paths.DATA_DIR}/processed_data.pkl",
'distance_function': cosine_distances,
}
###########################################################################
###########################################################################
func_test_fasttext(test_configs)
func_test_openai(test_configs)
contract_test_fasttext(test_configs)