SGSH/process_data.py at main · RUCKBReasoning/SGSH · GitHub

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

import os
import torch
import torch.nn as nn
import json
import pickle
import argparse
import numpy as np
from nltk import word_tokenize
from tqdm import tqdm
import re
import sys
import random
from transformers import BartTokenizer


def getNovelSubgraph(subkg, answer):
    g_nodes = subkg['g_node_names']
    g_edges = subkg['g_edge_types']
    g_adj = subkg['g_adj']
    g_edges = list(g_edges.values())
    relations = []
    seq = []
    for key, value in g_adj.items():
        subject = g_nodes[key]
        if subject == "none":
            continue
        else:
            for k in list(value.keys()):
                obj = g_nodes[k]
                if obj == "none" and k in list(g_adj.keys()):
                    value.update(g_adj[k])
    for key, value in g_adj.items():
        subject = g_nodes[key]
        if subject == "none":
            continue
        else:
            for k, relation in value.items():
                obj = g_nodes[k]
                if obj == "none":
                    continue
                else:
                    subject = subject.strip().lower()
                    obj = obj.strip().lower()
                    relation = relation.strip().lower()
                    relations.append(relation)
                    relation = relation.strip().split('/')[-1]
                    if relation.find('_')!=-1:
                        relation = relation.split('_')
                        relation = ' '.join(relation).strip()
                    fact = "<{}, {}, {}>".format(subject, relation, obj)
                    seq.append(fact)
    subkg = ", ".join(seq)
    return subkg, relations

def get_subkg_seq(subkg):
    seq = []
    maskList = []
    g_nodes = subkg['g_node_names']
    g_edges = subkg['g_edge_types']
    g_adj = subkg['g_adj']
    all_subjects = []
    all_objects = []
    relations = []
    for key, value in g_adj.items():
        subject = g_nodes[key]
        all_subjects.append(subject)
        for k, relation in value.items():
            obj = g_nodes[k]
            all_objects.append(obj)
            #####PQ relation is a list
            # relation = relation[0]
            # relations.append(relation)
            # if relation.find('/') >= 0:
            #     relation = relation.strip().split('/')[-1]

            #WQ relation is a str
            relation = relation.strip().split('/')[-1]

            if relation.find('_')!=-1:
                relation = relation.split('_')
                relation = ' '.join(relation).strip()
            fact = "<{}, {}, {}>".format(subject, relation, obj)
            seq.append(fact)
    subkg = ", ".join(seq)
    return subkg, relations


def encode_soft_skeleton_dataset(args, dataset, tokenizer, test, soft_tokens):
    max_seq_length = 496
    questions = []
    relations = []
    input_skeletons = []
    skeletons = []
    skeletons_train = []
    triple_nums = []
    answers = []
    subkgs = []
    answers_class = []
    count_answer = []
    for item in tqdm(dataset):
        question = item['outSeq']
        question = question.lower()
        questions.append(question)
        subkg = item['inGraph']
        answer = item['answers']
        answer = [ans.lower() for ans in answer]

        # subkg, relation = get_subkg_seq(subkg) #### PQ dataset
        subkg, node_names, relation, answer_class = getNovelSubgraph(subkg, answer) ####WQ dataset
        subkgs.append(subkg)

        if len(answer)>1:
            answer = [', '.join(answer)]
        if len(answer)==0:
            answer = ['']
        answers = answers + answer

        relations.append(relation)

    s = [i +' | ' + j for i, j in zip(subkgs, answers)] ### no relation

    with open(os.path.join(args.input_dir, 'train_skeleton.txt')) as f:
        for line in f.readlines():
            line = line.strip().lower()
            skeletons.append(line)

    input_ids = tokenizer.batch_encode_plus(s, max_length = max_seq_length, padding = "max_length", truncation = True, return_tensors="pt")

    input_ids['input_ids'] =  torch.cat([input_ids['input_ids'], torch.full((len(input_ids['input_ids']), soft_tokens), 50264)], 1)
    input_ids['attention_mask'] =  torch.cat([input_ids['attention_mask'], torch.full((len(input_ids['attention_mask']), soft_tokens), 1)], 1)

    source_ids = np.array(input_ids['input_ids'], dtype = np.int32)
    source_mask = np.array(input_ids['attention_mask'], dtype = np.int32)


    if not test:
        target_ids = tokenizer.batch_encode_plus(skeletons, max_length = max_seq_length, padding = "max_length", truncation = True, return_tensors="pt")
        target_ids = np.array(target_ids['input_ids'], dtype = np.int32)
    else:
        target_ids = np.array([], dtype = np.int32)

    return source_ids, source_mask, target_ids

def main():
    parser = argparse.ArgumentParser()
    parser.add_argument('--input_dir', required=True)
    parser.add_argument('--output_dir', required=True)
    parser.add_argument('--model_name_or_path', required=True)
    args = parser.parse_args()
    print('Loading!!!!')
    test_set = []
    train_set = []
    val_set = []

    with open(os.path.join(args.input_dir, 'test.json')) as f:
        for line in f.readlines():
            line = line.strip()
            line = json.loads(line)
            test_set.append(line)

    with open(os.path.join(args.input_dir, 'train.json')) as f:
        for line in f.readlines():
            line = json.loads(line.strip())
            train_set.append(line)
    with open(os.path.join(args.input_dir, 'dev.json')) as f:
        for line in f.readlines():
            line = json.loads(line.strip())
            val_set.append(line)
    if not os.path.isdir(args.output_dir):
        os.mkdir(args.output_dir)

    tokenizer = BartTokenizer.from_pretrained(args.model_name_or_path)
    for name, dataset in zip(('train', 'val', 'test'), (train_set, val_set, test_set)):
        print('Encode {} set'.format(name))
        outputs = encode_soft_skeleton_dataset(args, dataset, tokenizer,  name == 'test' or name == 'val', 16)
        print(type(outputs))
        print('shape of input_ids of questions, attention_mask of questions, input_ids of sparqls, choices and answers:')
        with open(os.path.join(args.output_dir, '{}.pt'.format(name)), 'wb') as f:
            for o in outputs:
                print(o.shape)
                pickle.dump(o, f)
if __name__ == '__main__':
    main()