PhaGCN/run_GCN.py at main · KennthShang/PhaGCN · 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
import  numpy as np

import  torch
from    torch import nn
from    torch import optim
from    torch.nn import functional as F


from    data import load_data, preprocess_features, preprocess_adj, sample_mask
import  model
from    config import  args
from    utils import masked_loss, masked_acc
import  pickle as pkl
import  scipy.sparse as sp

import random

seed = 123
np.random.seed(seed)
torch.random.manual_seed(seed)


if torch.cuda.is_available():
    torch.cuda.set_device(0)
else:
    print("Running with cpu")


adj        = pkl.load(open("Cyber_data/contig.graph",'rb'))
labels     = pkl.load(open("Cyber_data/contig.label",'rb'))
features   = pkl.load(open("Cyber_data/contig.feature",'rb'))
test_to_id = pkl.load(open("Cyber_data/contig.dict",'rb'))
idx_test   = pkl.load(open("Cyber_data/contig.mask",'rb'))

idx_test = np.array(idx_test)
labels = np.array(labels)

y_train = np.zeros(labels.shape)
y_test = np.zeros(labels.shape)


idx_train = np.array([i for i in range(len(labels)) if i not in idx_test])


train_mask = sample_mask(idx_train, labels.shape[0])
test_mask = sample_mask(idx_test, labels.shape[0])

y_train[train_mask] = labels[train_mask]
y_test[test_mask] = labels[test_mask]


features = sp.csc_matrix(features)

print('adj:', adj.shape)
print('features:', features.shape)
print('y:', y_train.shape, y_test.shape) # y_val.shape,
print('mask:', train_mask.shape, test_mask.shape) # val_mask.shape

features = preprocess_features(features) # [49216, 2], [49216], [2708, 1433]
supports = preprocess_adj(adj)

if torch.cuda.is_available():
    torch.cuda.set_device(0)
    device = torch.device('cuda')
    train_label = torch.from_numpy(y_train).long().to(device)
    num_classes = max(labels)+1
    train_mask = torch.from_numpy(train_mask.astype(np.bool)).to(device)
    test_label = torch.from_numpy(y_test).long().to(device)
    test_mask = torch.from_numpy(test_mask.astype(np.bool)).to(device)

    i = torch.from_numpy(features[0]).long().to(device)
    v = torch.from_numpy(features[1]).to(device)
    feature = torch.sparse.FloatTensor(i.t(), v, features[2]).float().to(device)

    i = torch.from_numpy(supports[0]).long().to(device)
    v = torch.from_numpy(supports[1]).to(device)
    support = torch.sparse.FloatTensor(i.t(), v, supports[2]).float().to(device)

else:
    train_label = torch.from_numpy(y_train).long()
    num_classes = max(labels)+1
    train_mask = torch.from_numpy(train_mask.astype(np.bool))
    test_label = torch.from_numpy(y_test).long()
    test_mask = torch.from_numpy(test_mask.astype(np.bool))

    i = torch.from_numpy(features[0]).long()
    v = torch.from_numpy(features[1])
    feature = torch.sparse.FloatTensor(i.t(), v, features[2]).float()

    i = torch.from_numpy(supports[0]).long()
    v = torch.from_numpy(supports[1])
    support = torch.sparse.FloatTensor(i.t(), v, supports[2]).float()

print('x :', feature)
print('sp:', support)
num_features_nonzero = feature._nnz()
feat_dim = feature.shape[1]


def accuracy(out, mask):
    pred = np.argmax(out, axis = 1)
    mask_pred = np.array([pred[i] for i in range(len(labels)) if mask[i] == True])
    mask_label = np.array([labels[i] for i in range(len(labels)) if mask[i] == True])
    return np.sum(mask_label == mask_pred)/len(mask_pred)

net = model.GCN(feat_dim, num_classes, num_features_nonzero)
if torch.cuda.is_available():
    net.to(device)
optimizer = optim.Adam(net.parameters(), lr=0.01)#args.learning_rate


_ = net.train()
for epoch in range(args.epochs*2):
    # forward pass
    out = net((feature, support))
    #out = out[0]
    loss = masked_loss(out, train_label, train_mask)
    loss += args.weight_decay * net.l2_loss()
    # backward pass
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()
    # output
    if epoch % 10 == 0:
        # calculating the acc
        _ = net.eval()
        out = net((feature, support))
        if torch.cuda.is_available():
            acc_train = accuracy(out.detach().cpu().numpy(), train_mask.detach().cpu().numpy())
        else:
            acc_train = accuracy(out.detach().numpy(), train_mask.detach().numpy())
        #acc_test = accuracy(out.detach().cpu().numpy(), test_mask.detach().cpu().numpy())
        print(epoch, loss.item(), acc_train)
        if acc_train > 0.978:
            break
    _ = net.train()


net.eval()
out = net((feature, support))
out = F.softmax(out,dim =1)
if torch.cuda.is_available():
    out = out.cpu().detach().numpy()
else:
    out = out.detach().numpy()

pred = np.argmax(out, axis = 1)
score = np.max(out, axis = 1)

mode = "testing"
if mode == "validation":
    print(classification_report(labels, pred))
    print(accuracy(out, train_mask.detach().cpu().numpy()))
    mask = test_mask.detach().cpu().numpy()
    test_pred = np.array([pred[i] for i in range(len(pred)) if mask[i] == True])
    test_label = np.array([labels[i] for i in range(len(labels)) if mask[i] == True])
    print(classification_report(test_label, test_pred))
    print(np.sum(test_label == test_pred)/len(test_pred))


pred_to_label = {0:"Ackermannviridae", 1:"Autographiviridae", 2:"Demerecviridae",
3:"Drexlerviridae", 4:"Herelleviridae", 5:"Myoviridae", 6:"Podoviridae", 7:"Siphoviridae"}


with open("prediction.csv", 'w') as f_out:
    _ = f_out.write("contig_names,prediction,score\n")
    for key in test_to_id.keys():
        if labels[test_to_id[key]] == -1:
            _ = f_out.write(str(key) + "," + str(pred_to_label[pred[test_to_id[key]]]) + "," + str(score[test_to_id[key]]) + "\n")
        else:
            _ = f_out.write(str(key) + "," + str(pred_to_label[labels[test_to_id[key]]]) + "," + str(1) + "\n")