-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathRUE_build.py
198 lines (177 loc) · 11.6 KB
/
RUE_build.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
import time
from utils.utils import Tools, FilePathBuilder
from utils.vector_utils import BagOfWordsEmbedding, UniXcoderEmbedding
from utils.rerank_utils import SingleReranking
device = "cpu"
class RUEPromptBuilder:
def __init__(self, benchmark, repo_base_dir, reranker_type, repos, tasks, tokenizer):
self.benchmark = benchmark
self.repo_base_dir = repo_base_dir
if reranker_type == 'bow':
self.vector_builder = BagOfWordsEmbedding()
elif reranker_type == 'unixcoder':
self.vector_builder = UniXcoderEmbedding(device=device)
else:
raise NotImplementedError
self.reranker = SingleReranking(self.vector_builder)
self.repos = repos
self.tasks = tasks
self.tokenizer = tokenizer
def build_prompt(self, output_file):
new_prompt_lines = []
task_cnt = 0
start_time = time.time()
max_prompt_token = 0
for repo in self.repos:
repo_method_input_file = FilePathBuilder.repo_methods_path(repo, self.benchmark)
repo_embedding_methods_file = self.vector_builder.vector_file_path(repo_method_input_file)
repo_embedding_methods = Tools.load_pickle(repo_embedding_methods_file)
repo_type_input_file = FilePathBuilder.repo_types_path(repo, self.benchmark)
repo_embedding_types_file = self.vector_builder.vector_file_path(repo_type_input_file)
repo_embedding_types = Tools.load_pickle(repo_embedding_types_file)
repo_type_lines = Tools.load_pickle(repo_type_input_file)
repo_type_name_docstrings = []
repo_type_methods = {}
for key, methods in repo_type_lines.items():
class_name, relative_path = key
if class_name not in repo_type_methods:
repo_type_name_docstrings.append(class_name)
repo_type_methods[class_name] = []
repo_type_methods[class_name].extend(methods)
repo_method_name_docstrings = []
for line in repo_embedding_methods:
repo_method_name_docstrings.append(line['metadata']['name'])
batch_size = 64
for task in self.tasks:
print(task['metadata']['task_id'])
if task['metadata']['task_id'].split('/')[0] == repo:
task_cnt += 1
print('Re-ranking for task: ', task['metadata']['task_id'], 'Task count: ', task_cnt)
score_candidate_methods = []
for top_k_signature in task['metadata']['top_k_context']:
score_candidate_methods.append({
'fpath_tuple': top_k_signature['fpath_tuple'],
'method': top_k_signature['method'],
'sim_score': top_k_signature['sim_score'],
'start_line_no': top_k_signature['start_line_no'],
'end_line_no': top_k_signature['end_line_no'],
'type_query': 'signature'
})
prediction_samples = task['choices']
clean_prediction = Tools.clean_output(prediction_samples[0]['text'])
extracted_types = task['prediction.types']
extracted_methods = task['prediction.methods']
# clean_prediction = Tools.clean_output(prediction_samples)
# extracted_types = task['ground_truth.types']
# extracted_methods = task['ground_truth.methods']
print('Start reranking for extracted types for task: ', task['metadata']['task_id'])
candidate_method_names = []
# new_extracted_types = []
for extracted_type in extracted_types:
doc_embeddings = []
docs = []
for line in repo_embedding_types:
docs.append(line)
doc_embeddings.append(line['data'][0]['name_embedding'])
top_k_extracted_types = self.reranker.rerank_batch(extracted_type, docs, doc_embeddings, 1)
max_score_retreived_type = top_k_extracted_types[0][0]['name']
print(extracted_type, max_score_retreived_type, top_k_extracted_types[0][1])
candidate_method_names.extend(repo_type_methods[max_score_retreived_type])
print('Start reranking for extracted methods for task: ', task['metadata']['task_id'])
for extracted_method in extracted_methods:
doc_embeddings = []
docs = []
for line in repo_embedding_methods:
docs.append(line)
doc_embeddings.append(line['data'][0]['name_embedding'])
top_k_extracted_methods = self.reranker.rerank_batch(extracted_method, docs, doc_embeddings, 1)
max_score_retreived_method = top_k_extracted_methods[0][0]['metadata']['name']
print(extracted_method, max_score_retreived_method, top_k_extracted_methods[0][1])
candidate_method_names.append(max_score_retreived_method)
candidate_methods = []
print('Start retrieve candidate methods for task: ', task['metadata']['task_id'])
print('Number of repo methods: ', len(repo_embedding_methods))
print('Number of candidate method names: ', len(candidate_method_names))
for repo_method in repo_embedding_methods:
if not Tools.is_finding_method(task, repo_method):
for candidate_method_name in candidate_method_names:
if candidate_method_name in repo_method['metadata']['body_raw']:
candidate_methods.append(repo_method)
break
print('Number of candidate methods: ', len(candidate_methods))
if len(candidate_methods) > 0:
doc_embeddings = []
docs = []
for line in candidate_methods:
if not Tools.is_finding_method(task, line):
docs.append(line)
doc_embeddings.append(line['data'][0]['body_embedding'])
top_k_draft_methods = self.reranker.rerank_batch(clean_prediction, docs, doc_embeddings, 20)
print('Query: ', clean_prediction)
top_method_query = top_k_draft_methods[0][0]
print('Top similar draft method: ', top_method_query['metadata']['body_raw'])
print('Score: ', top_k_draft_methods[0][1])
for top_k_draft_method in top_k_draft_methods:
score_candidate_methods.append({
'fpath_tuple': top_k_draft_method[0]['metadata']['fpath_tuple'],
'method': top_k_draft_method[0]['metadata']['name'],
'sim_score': top_k_draft_method[1],
'start_line_no': top_k_draft_method[0]['metadata']['start_line_no'],
'end_line_no': top_k_draft_method[0]['metadata']['end_line_no'],
'type_query': 'draft'
})
score_candidate_methods = sorted(score_candidate_methods, key=lambda x: x['sim_score'], reverse=True)
print('Start build prompt for extracted methods for task: ', task['metadata']['task_id'])
prepend_context = "// Here are some relevant code fragments from other files of the repo:\n"
seperator = '// ' + '-' * 50
prepend_context += seperator + '\n'
context_class = task['metadata']['left_context'] + '<FILL_FUNCTION_BODY>' + task['metadata']['right_context']
# context_class = task['prompt'].split('Based on above, complete the method body of the class\n\n')[-1]
len_current_token = self.tokenizer(prepend_context + context_class + task['metadata']['ground_truth'], return_tensors="pt")['input_ids'].size()[1]
chosen_context = []
for score_method in score_candidate_methods[:20]:
# content = repo_embedding_methods[doc_id]
f_path = '/'.join(score_method['fpath_tuple'])
f_paths_str = f'// {f_path}'
left_content = Tools.get_lcontext_method(score_method, self.repo_base_dir)
left_content_lines = left_content.split('\n')
left_content_lines_comment = [f'// {line}' for line in left_content_lines]
f_path_comment = f'// the below code fragment can be found in:'
block_str = '\n'.join([f_path_comment, f_paths_str, seperator] + left_content_lines_comment + [seperator]) + '\n'
# block_str = '\n'.join(left_content_lines_comment + [seperator]) + '\n'
block_token_len = self.tokenizer(block_str, return_tensors="pt")['input_ids'].size()[1]
if len_current_token + block_token_len > 8000:
break
prepend_context += block_str
len_current_token += block_token_len
chosen_context.append(
(score_method['fpath_tuple'],
left_content,
score_method['sim_score'],
score_method['method'],
score_method['start_line_no'],
score_method['end_line_no'],
score_method['type_query']),
)
prepend_context += """// Based on above, complete the method body of the class\n"""
if len(chosen_context) > 0:
prompt = prepend_context + '\n' + context_class
else:
prompt = context_class
max_prompt_token = max(max_prompt_token, self.tokenizer(prompt, return_tensors="pt")['input_ids'].size()[1])
new_prompt_line = task.copy()
new_prompt_line['prompt'] = prompt
new_prompt_line['metadata']['top_k_context'] = [
{
'fpath_tuple': x[0],
'left_content': x[1],
'sim_score': x[2],
'method': x[3],
'start_line_no': x[4],
'end_line_no': x[5],
'type_query': x[6]
} for x in chosen_context
]
new_prompt_lines.append(new_prompt_line)
Tools.dump_jsonl(new_prompt_lines, output_file)
print('Total time: ', time.time() - start_time)