GNN4CodeClassification/model_test.py at main · yizhidou/GNN4CodeClassification · 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
from typing import List
import os
import numpy as np
from utils import *
from loss import euclidean_distance
import random, json, argparse
from graph_embedding_calculation import GraphEmbeddingCalculaor


class TestTaskBasic: # base class for multiple testing tasks
    def __init__(self, model_params_savepath, device):
        self.device = device
        self.model_params_savepath = model_params_savepath
        with open(model_params_savepath) as params_loader:
            self.model_params = json.load(params_loader)
        self.model_id = model_params_savepath.rsplit('/')[-1][:-5]
        # self.model_indicator = self.ckpt_path.rsplit('/', 1)[-1][:-5]
        self.nm_dir = self.model_params['dataset_params']['nm_dir']
        self.am_dir = self.model_params['dataset_params']['am_dir']
        self.dataset_name = self.model_params['dataset_params']['dataset_name']
        self.label_loader = LabelLoader(dataset_name=self.dataset_name,
                                        cwe_to_label_mapping_path=self.model_params['dataset_params'][
                                            'cwe_to_label_mapping_record'])
        with open(self.model_params['dataset_params']['train_vali_test_divide_record']) as divide_loader:
            self.train_vali_test_divide_record = json.load(divide_loader)
        # self.sample_fn_list_under_test = self.train_vali_test_divide_record['test_sha']
        self.ckpt_path = os.path.join(self.model_params['ckpt_save_dir'], self.model_id + '.ckpt')

        graph_embedding_record_savedir = ''
        if not os.path.isdir(graph_embedding_record_savedir):
            os.system('mkdir -p {}'.format(graph_embedding_record_savedir))
        self.graph_embedding_record_savepath = os.path.join(graph_embedding_record_savedir,
                                                            self.model_id + '_embeddings.json')
        if self.model_params['graph_embedding_net_settings']['prop_type'] == 'embedding':
            print('prop type is embedding, so graph embeddings could be calculated first~')
            if os.path.isfile(self.graph_embedding_record_savepath):
                print('embeddings found! yeah~')
            else:
                print('{} does not exit~ so we calculate them for future use~'.format(
                    self.graph_embedding_record_savepath))
                GraphEmbeddingCalculaor(model_params_path=model_params_savepath,
                                        record_savepath=self.graph_embedding_record_savepath,
                                        device=self.device).dump_to_record()
                print('embedding calculation has done~')
            with open(self.graph_embedding_record_savepath) as json_loader:
                self.sample_fn_to_graph_embedding_dict = json.load(json_loader)
        else:
            assert self.model_params['graph_embedding_net_settings']['prop_type'] == 'matching'
            print('prop type is matching, so there is no graph embeddings reserved')
            self.sample_fn_to_graph_embedding_dict = None

        # self.model = self.get_loaded_model_from_ckpt()  # actually, you don't even need to load a model if all embeddings are avaliable

    def get_loaded_model_from_ckpt(self):
        model, _ = build_model(self.model_params)
        checkpoint = torch.load(self.ckpt_path, map_location=self.device)
        model.load_state_dict(checkpoint['model_state_dict'])
        model.to(self.device)
        model.eval()
        return model

    def get_graph_embedding_for_one_sample(self, sample_sha):
        sample_nm_path = os.path.join(self.nm_dir, sample_sha + '.nm')
        sample_am_path = os.path.join(self.am_dir, sample_sha + '.am')
        sample_nm = np.loadtxt(sample_nm_path, delimiter=',').astype(np.float32)
        sample_am = np.loadtxt(sample_am_path, delimiter=',').astype(int)
        edge_feature = np.zeros(shape=(sample_am.shape[0], 13), dtype=np.float32)
        from_idx = sample_am[:, 0]
        to_idx = sample_am[:, 1]
        graph_idx = np.zeros(shape=(sample_nm.shape[0],), dtype=np.int64)

        # model.to(self.device)
        # model.eval()
        with torch.no_grad():
            sample_graph_embedding = self.model(torch.from_numpy(sample_nm).to(self.device),
                                                torch.from_numpy(edge_feature).to(self.device),
                                                torch.from_numpy(from_idx).to(self.device),
                                                torch.from_numpy(to_idx).to(self.device),
                                                torch.from_numpy(graph_idx).to(self.device), 1)
        return sample_graph_embedding

    def get_graph_embeddings_for_samples(self, sample_fn_list):
        nm_list = []
        am_list = []
        exist_sample_fn_list = []
        for sample_sha in sample_fn_list:
            sample_nm_path = os.path.join(self.nm_dir, sample_sha + '.nm')
            sample_am_path = os.path.join(self.am_dir, sample_sha + '.am')
            if not os.path.isfile(sample_nm_path):
                continue
            sample_nm = np.loadtxt(sample_nm_path, delimiter=',').astype(np.float32)
            sample_am = np.loadtxt(sample_am_path, delimiter=',').astype(int)
            nm_list.append(sample_nm)
            am_list.append(sample_am)
            exist_sample_fn_list.append(sample_sha)
        packed_data = pack_datapoint(nm_list=nm_list,
                                     am_list=am_list)
        node_features, edge_features, from_idx, to_idx, graph_idx = get_graph(packed_data)
        with torch.no_grad():
            graph_embeddings = self.model(node_features.to(self.device), edge_features.to(self.device),
                                          from_idx.to(self.device),
                                          to_idx.to(self.device),
                                          graph_idx.to(self.device), len(nm_list))
        # print('length of sample_{} = {}'.format(sample_fn_list, euclidean_distance(graph_embeddings, 0)))
        return graph_embeddings, np.array(exist_sample_fn_list, dtype=object)

    def batched_graph_embedding_generator(self, batch_size):
        batched_sample_fn_list = []
        batch_idx = 0
        if self.sample_fn_to_graph_embedding_dict:
            batched_graph_embedding_list = []
            for sample_sha in self.sample_fn_list_under_test:
                batch_idx += 1
                if len(batched_sample_fn_list) % batch_size == 0 and len(batched_sample_fn_list) > 0:
                    graph_embeddings_of_one_batch = torch.from_numpy(np.stack(batched_graph_embedding_list)).to(
                        self.device)
                    yield graph_embeddings_of_one_batch, np.array(batched_sample_fn_list, dtype=object)
                    batched_sample_fn_list = []
                    batched_graph_embedding_list = []
                else:
                    if not sample_sha in self.sample_fn_to_graph_embedding_dict:
                        continue
                    batched_sample_fn_list.append(sample_sha)
                    batched_graph_embedding_list.append(np.array(self.sample_fn_to_graph_embedding_dict[sample_sha]))
            if len(batched_sample_fn_list) > 0:
                graph_embeddings_of_last_batch = torch.from_numpy(np.stack(batched_graph_embedding_list)).to(
                    self.device)
                yield graph_embeddings_of_last_batch, np.array(batched_sample_fn_list, dtype=object)
        else:
            for sample_sha in self.sample_fn_list_under_test:
                batch_idx += 1
                if len(batched_sample_fn_list) % batch_size == 0 and len(batched_sample_fn_list) > 0:
                    # print('{}: embedding of {} is generating...'.format(sample_fn_idx, batched_sample_fn_list))
                    yield self.get_graph_embeddings_for_samples(sample_fn_list=batched_sample_fn_list)
                    batched_sample_fn_list = []
                else:
                    batched_sample_fn_list.append(sample_sha)
            # print('the num of samples of type_{} is: {}'.format(debug_label, sample_fn_idx))
            if len(batched_sample_fn_list) > 0:
                yield self.get_graph_embeddings_for_samples(sample_fn_list=batched_sample_fn_list)

    def batched_graph_generator(self, sample_fn_list, batch_size):
        batch_sample_fn_list = []
        batch_nm_list = []
        batch_am_list = []
        for sample_sha in sample_fn_list:
            sample_nm_path = os.path.join(self.nm_dir, sample_sha + '.nm')
            sample_am_path = os.path.join(self.am_dir, sample_sha + '.am')
            if not os.path.isfile(sample_nm_path):
                continue
            try:
                sample_nm = np.loadtxt(sample_nm_path, delimiter=',').astype(np.float32)
            except ValueError:
                continue
            sample_am = np.loadtxt(sample_am_path, delimiter=',').astype(int)
            batch_nm_list.append(sample_nm)
            batch_am_list.append(sample_am)
            batch_sample_fn_list.append(sample_sha)
            if len(batch_sample_fn_list) == batch_size:
                # packed_data = pack_datapoint(nm_list=batch_nm_list,
                #                              am_list=batch_am_list)
                yield batch_nm_list, batch_am_list, batch_sample_fn_list
                batch_sample_fn_list = []
                batch_nm_list = []
                batch_am_list = []