def return_scores_and_labels_for_5fold(i, type, model_number, projectName):
# 有kmer的
# load datset
# print('load dataset with kmer')
datset_test = LncRNA_Protein_Interaction_dataset_1hop_1220_InMemory(
f'data/dataset/{projectName}_inMemory{type}_test_{i}')
print(f'载入数据集：data/dataset/{projectName}_inMemory{type}_test_{i}')
test_loader = DataLoader(datset_test, batch_size=60)
# load model
# print('load model with kmer')
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = Net_1(datset_test.num_node_features)
model.load_state_dict(
torch.load(
f'result/{projectName}{type}/model_{i}_fold/{model_number}'))
print(f'载入模型：result/{projectName}{type}/model_{i}_fold/{model_number}')
# run test
scores_temp, labels_temp = return_scores_and_labels(
model, test_loader, device)
return scores_temp, labels_temp
parser.add_argument('--testName', help='name of this test')
parser.add_argument('--path_model', help='the path of model')
parser.add_argument('--path_dataset', help='the path of test dataset')
return parser.parse_args()
if __name__ == "__main__":
args = parse_args()
# load datset
print('load dataset')
datset_test = LncRNA_Protein_Interaction_dataset(args.path_dataset)
test_loader = DataLoader(datset_test, batch_size=60)
# load model
print('load model')
device = torch.device('cuda')
model = Net_1(datset_test.num_node_features).to(device)
model.load_state_dict(torch.load(args.path_model))
# run test
print('run test')
Accuracy, Precision, Sensitivity, Specificity, MCC = Accuracy_Precision_Sensitivity_Specificity_MCC(
model, test_loader, device)
output = 'Accuracy: {:.5f}, Precision: {:.5f}, Sensitivity: {:.5f}, Specificity: {:.5f}, MCC: {:.5f}'.format(
Accuracy, Precision, Sensitivity, Specificity, MCC)
print(output)
# output log
path_log_folder = osp.join(r'.\result', args.testName)
if not osp.exists(path_log_folder):
os.mkdir(path_log_folder)
path_log = osp.join(path_log_folder, 'log.txt')
with open(path_log, 'w') as log:
log.write(output)
Accuracy_list = []
Precision_list = []
Sensitivity_list = []
Specificity_list = []
MCC_list = []
for i in range(k):
print('{:d}-fold start'.format(i+1))
# 创建保存模型的文件夹
os.makedirs(saving_path + f'/model_{i}_fold')
# 创建模型
num_of_classes = 2
if i != 0:
del model
gc.collect()
model = Net_1(dataset.num_node_features, num_of_classes).to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=LR, weight_decay=L2_weight_decay)
# scheduler = lr_scheduler.MultiStepLR(optimizer,milestones=[int(num_of_epoch * 0.2),int(num_of_epoch * 0.8)],gamma = 0.8)
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer=optimizer, gamma=0.95)
# 训练集和测试集
test_dataset_start = 0 + i * step_length
test_dataset_end = (i + 1) * step_length
test_dataset = dataset[test_dataset_start:test_dataset_end]
train_dataset = dataset[0:test_dataset_start] + dataset[test_dataset_end:dataset.len()]
print('number of samples in testing dataset：', len(test_dataset), 'number of samples in training dataset：', len(train_dataset))
print('training dataset')
dataset_analysis(train_dataset)
print('testing dataset')
read_random_node_embedding()
# load k-mer
if args.noKmer == 0:
lncRNA_3_mer_path = f'data/lncRNA_3_mer/{args.interactionDatasetName}/lncRNA_3_mer.txt'
protein_2_mer_path = f'data/protein_2_mer/{args.interactionDatasetName}/protein_2_mer.txt'
load_node_k_mer(lncRNA_list, 'lncRNA', lncRNA_3_mer_path)
load_node_k_mer(protein_list, 'protein', protein_2_mer_path)
# 执行检查
load_exam(args.noKmer, lncRNA_list, protein_list)
# 载入模型
device = torch.device('cuda')
model = Net_1(178).to(device)
model.load_state_dict(torch.load(args.modelPath))
# 开始case study
# 构建从serial number到名字的字典
dict_interactionKey_interaction = return_dict_interactionKey_interaction(interaction_list)
# 创建保存数据集和log的文件夹
case_study_path = f'data/case_study/{args.caseStudyName}/datasets'
if not osp.exists(case_study_path):
os.makedirs(case_study_path)
print(f'创建了文件夹：{case_study_path}')
log_path = f'data/case_study/{args.caseStudyName}/logs'
if not osp.exists(log_path):
os.makedirs(log_path)