-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathbuild_instructions.py
141 lines (111 loc) · 7.62 KB
/
build_instructions.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
import os
import json
from data_manager import DataManager
from tqdm import tqdm
import argparse
from prompt_templates import SUBGRAPH_REASON_PROMPT, NEIGHBOR_REASON_PROMPT, CLOSE_PATH_REASON_PROMPT
def close_path_finder(data_manager:DataManager, triple):
head, relation, tail = triple
close_paths = list(data_manager.bfs_paths(head, tail))
if close_paths:
path_degrees = []
for path in close_paths:
degree_sum = sum(data_manager.relation_degree_dict[rel] for _, rel, _ in path)
path_degrees.append((degree_sum, path))
path_degrees.sort(key=lambda x: x[0])
top_paths = [path for _, path in path_degrees[:data_manager.max_reason_paths]]
top_paths.reverse()
return top_paths
return []
def build_instructions(dataset, train_size, subgraph_type, neg_num, version):
setting = "transductive"
data_manager = DataManager(dataset=dataset, setting=setting, train_size=train_size)
paths_dir = f"instructions{version}/{dataset}"
os.makedirs(paths_dir, exist_ok=True)
sft_instructions = []
for pos_triple in tqdm(data_manager.path_set, desc=f"Processing {dataset} - setting: {setting} - Train_size: {train_size}"):
pos_head, relation, pos_tail = pos_triple
removed_from_head = (relation, pos_tail, 1)
removed_from_tail = (relation, pos_head, -1)
data_manager.entity2relationtail_dict[pos_head].remove(removed_from_head)
data_manager.entity2relationtail_dict[pos_tail].remove(removed_from_tail)
pos_type_prompt = data_manager.build_type_prompt(pos_triple)
sft_instructions.append({"instruction": pos_type_prompt, "input": "", "output": "Y"})
if subgraph_type == "combine":
pos_neighbor_triples = data_manager.neighbor_triple_finder(pos_triple)
pos_close_paths = close_path_finder(data_manager, pos_triple)
pos_reasoning_paths = "\n".join(
" -> ".join(data_manager.triple_to_sentence(triple) for triple in path)
for path in pos_close_paths
)
pos_subgraph_prompt = SUBGRAPH_REASON_PROMPT.format(neighbor_triples="\n".join(pos_neighbor_triples), reasoning_paths=pos_reasoning_paths, test_triple=data_manager.triple_to_sentence(pos_triple))
elif subgraph_type == "neighbor-only":
pos_neighbor_triples = data_manager.neighbor_triple_finder(pos_triple)
pos_subgraph_prompt = NEIGHBOR_REASON_PROMPT.format(neighbor_triples="\n".join(pos_neighbor_triples), test_triple=data_manager.triple_to_sentence(pos_triple))
elif subgraph_type == "path-only":
pos_close_paths = close_path_finder(data_manager, pos_triple)
pos_reasoning_paths = "\n".join(
" -> ".join(data_manager.triple_to_sentence(triple) for triple in path)
for path in pos_close_paths
)
pos_subgraph_prompt = CLOSE_PATH_REASON_PROMPT.format(reasoning_paths=pos_reasoning_paths, test_triple=data_manager.triple_to_sentence(pos_triple))
sft_instructions.append({"instruction": pos_subgraph_prompt, "input": "", "output": "Y"})
neg_samples = data_manager.neg_sampling(pos_triple, neg_num)
for neg_triple in neg_samples:
neg_type_prompt = data_manager.build_type_prompt(neg_triple)
sft_instructions.append({"instruction": neg_type_prompt, "input": "", "output": "N"})
if subgraph_type == "combine":
neg_neighbor_triples = data_manager.neighbor_triple_finder(neg_triple)
neg_close_paths = close_path_finder(data_manager, neg_triple)
neg_reasoning_paths = "\n".join(
" -> ".join(data_manager.triple_to_sentence(triple) for triple in path)
for path in neg_close_paths
)
neg_subgraph_prompt = SUBGRAPH_REASON_PROMPT.format(neighbor_triples="\n".join(neg_neighbor_triples), reasoning_paths=neg_reasoning_paths, test_triple=data_manager.triple_to_sentence(neg_triple))
elif subgraph_type == "neighbor-only":
neg_neighbor_triples = data_manager.neighbor_triple_finder(neg_triple)
neg_subgraph_prompt = NEIGHBOR_REASON_PROMPT.format(neighbor_triples="\n".join(neg_neighbor_triples), test_triple=data_manager.triple_to_sentence(neg_triple))
elif subgraph_type == "path-only":
neg_close_paths = close_path_finder(data_manager, neg_triple)
neg_reasoning_paths = "\n".join(
" -> ".join(data_manager.triple_to_sentence(triple) for triple in path)
for path in neg_close_paths
)
neg_subgraph_prompt = CLOSE_PATH_REASON_PROMPT.format(reasoning_paths=neg_reasoning_paths, test_triple=data_manager.triple_to_sentence(neg_triple))
sft_instructions.append({"instruction": neg_subgraph_prompt, "input": "", "output": "N"})
data_manager.entity2relationtail_dict[pos_head].append(removed_from_head)
data_manager.entity2relationtail_dict[pos_tail].append(removed_from_tail)
sft_instructions_path = f"{paths_dir}/{dataset}_train_size_{train_size}_{subgraph_type}.json"
with open(sft_instructions_path, "w", encoding="utf-8") as f:
json.dump(sft_instructions, f, ensure_ascii=False, indent=4)
def build_vanilla_instructions(dataset, train_size, neg_num, version):
setting = "transductive" # 指令构建默认是transductive，用训练集
data_manager = DataManager(dataset=dataset, setting=setting, train_size=train_size)
paths_dir = f"vanilla_instructions{version}/{dataset}"
os.makedirs(paths_dir, exist_ok=True)
sft_instructions = []
for pos_triple in tqdm(data_manager.path_set, desc=f"Processing {dataset} - setting: {setting} - Train_size: {train_size}"):
pos_vanilla_prompt = data_manager.build_vanilla_prompt(pos_triple)
sft_instructions.append({"instruction": pos_vanilla_prompt, "input": "", "output": "Y"})
neg_samples = data_manager.neg_sampling(pos_triple, neg_num)
for neg_triple in neg_samples:
neg_vanilla_prompt = data_manager.build_vanilla_prompt(neg_triple)
sft_instructions.append({"instruction": neg_vanilla_prompt, "input": "", "output": "N"})
sft_instructions_path = f"{paths_dir}/{dataset}_train_size_{train_size}.json"
with open(sft_instructions_path, "w", encoding="utf-8") as f:
json.dump(sft_instructions, f, ensure_ascii=False, indent=4)
def main():
parser = argparse.ArgumentParser(description='Process datasets with given hyperparameters')
parser.add_argument("--dataset", type=str, choices=["FB15k-237-subset", "NELL-995-subset", "WN18RR-subset"], default="FB15k-237-subset")
parser.add_argument("--train_size", type=str, choices=["full", "1000", "2000"], default="full", help="Size of the training data")
parser.add_argument("--neg_num", type=int, default=3, help="Number of negative samples")
parser.add_argument("--prompt_type", type=str, default="CATS", choices=["CATS", "vanilla"])
parser.add_argument("--subgraph_type", type=str, default="combine", choices=["neighbor-only", "path-only", "combine"])
parser.add_argument("--version", type=str, default="")
args = parser.parse_args()
if args.prompt_type == "CATS":
build_instructions(args.dataset, args.train_size, args.subgraph_type, args.neg_num, args.version)
elif args.prompt_type == "vanilla":
build_vanilla_instructions(args.dataset, args.train_size, args.neg_num, args.version)
if __name__ == "__main__":
main()