-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.py
136 lines (128 loc) · 6.15 KB
/
train.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
import os
import numpy as np
import pandas as pd
import torch
import torch.nn as nn
import torch.optim as optim
import glob
import argparse
import ast
import time
import utils
from datetime import datetime
import json
from pdb import set_trace
from multiprocessing import Process, Manager
from MNPWAD import MNPWAD
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument('--datapath', type=str, default='data', help='data path')
parser.add_argument('--datasets', type=str, default='',
help='dataset name of the path')
parser.add_argument('--outputpath', type=str, default='output')
parser.add_argument('--algo', type=str, default="MNPWAD")
parser.add_argument('--trainflag', type=str, default="")
parser.add_argument('--labeled_ratio', type=float, default=0.01)
parser.add_argument('--runs', type=int, default=1)
parser.add_argument('--use_es', type=ast.literal_eval, default=True)
parser.add_argument('--base_model', type=str, default='MNPWAD')
parser.add_argument('--gpu', type=str, default='0')
parser.add_argument('--batch_size', type=int, default=128)
parser.add_argument('--lr', type=float, default=0.005)
parser.add_argument('--n_emb', type=int, default=8)
parser.add_argument('--m1', type=float, default=0.02)
parser.add_argument('--debug', type=ast.literal_eval, default=False)
parser.add_argument('--pretrainAE', type=ast.literal_eval, default=True)
parser.add_argument('--n_prototypes', type=int, default=0)
parser.add_argument('--dataset2n_prototypes', type=str, default='')
return parser.parse_args()
def train_model(dataset_name,device,args,results,output_path):
f = glob.glob(os.path.join(args.datapath, f'{dataset_name}.csv'))
assert len(f) == 1
df = pd.read_csv(f[0])
logger=utils.logger(f'{output_path}/log/{dataset_name}.txt')
model_name = args.algo
logger.info("------------------ Dataset: [%s] ------------- Device: [%s]-------------" % (dataset_name,device),print_text=True)
df.replace([np.inf, -np.inf], np.nan, inplace=True)
df.fillna(method='ffill', inplace=True)
x = df.values[:, :-1]
y = np.array(df.values[:, -1], dtype=int)
x_train, y_train, x_test, y_test, x_val, y_val = utils.split_train_test_val(x, y,
test_ratio=0.35,
val_ratio=0.05,
random_state=2024,
del_features=True)
if args.labeled_ratio<1:
semi_y = utils.semi_setting_ratio(y_train, labeled_ratio=args.labeled_ratio)
else:
semi_y=y_train
runs=args.runs
rauc, raucpr, rtime = np.zeros(runs), np.zeros(runs), np.zeros(runs)
params={'device':device,'use_es':args.use_es,'logger':logger,'base_model':args.base_model,'output_path':output_path,'dataset_name':dataset_name
,'batch_size':args.batch_size,'n_emb':args.n_emb,'lr':args.lr,'m1':args.m1,'lambda_kl':args.lambda_kl,'n_prototypes':args.n_prototypes}
if args.dataset2n_prototypes!='':
params['n_prototypes']=utils.get_n_prototypes(dataset_name,args.dataset2n_prototypes)
for i in range(runs):
st = time.time()
params['seed'] = 42+i
model = eval(model_name)(**params)
model.fit(x_train, semi_y, val_x=x_val, val_y=y_val,pretrainAE=args.pretrainAE)
score = model.predict(x_test)
auroc, aupr = utils.evaluate(y_test, score)
rtime[i] = time.time() - st
rauc[i] = auroc
raucpr[i] = aupr
txt = f'{dataset_name}, AUC-ROC: {auroc:.3f}, AUC-PR: {aupr:.3f}, ' \
f'time: {rtime[i]:.1f}, runs: [{i+1}/{runs}]'
logger.info(txt,print_text=False)
doc = open(output_path+'/middle_result/' + f'{dataset_name}.csv', 'a')
print(txt, file=doc)
doc.close()
print_text = f"{dataset_name}, AUC-ROC, {np.average(rauc):.3f}±{np.std(rauc):.3f}," \
f" AUC-PR, {np.average(raucpr):.3f}±{np.std(raucpr):.3f}, {np.average(rtime):.1f}," \
f" {runs}runs, {model.n_prototypes}_prototypes," \
f" {model_name}, {str(model.param_lst)}"
logger.info(print_text,print_text=True)
doc = open(output_path+'/middle_result/' + f'{dataset_name}.csv', 'a')
print(print_text, file=doc)
doc.close()
results[dataset_name] = (print_text,np.average(rauc),np.average(raucpr))
if __name__ == '__main__':
args=parse_args()
current_date = datetime.now()
date_string = current_date.strftime('%Y%m%d%H%M')
output_path=os.path.join(args.outputpath, args.trainflag+'-'+date_string)
if not os.path.exists(output_path):
os.makedirs(output_path)
os.makedirs(os.path.join(output_path,'log'), exist_ok=True)
os.makedirs(os.path.join(output_path,'checkpoints'), exist_ok=True)
os.makedirs(os.path.join(output_path,'middle_result'), exist_ok=True)
torch.multiprocessing.set_start_method('spawn')
datasets = args.datasets.split(',')
if torch.cuda.is_available():
gpu_list=args.gpu.split(',') if ',' in args.gpu else [args.gpu]
manager = Manager()
results = manager.dict()
processes = []
for i, dataset in enumerate(datasets):
device= 'cuda:'+gpu_list[i%len(gpu_list)] if torch.cuda.is_available() else 'cpu'
p = Process(target=train_model, args=(dataset, device,args, results,output_path))
processes.append(p)
p.start()
if args.debug:
break
for p in processes:
p.join()
results = dict(results)
summary = []
doc = open(output_path+'/all_result.csv', 'a')
logger=utils.logger(f'{output_path}/all_result.txt')
for dataset in datasets:
print(results[dataset][0], file=doc)
summary.append(results[dataset][1:])
logger.info(results[dataset][0],print_text=True)
text = f"avg, AUC-ROC, {np.average([x[0] for x in summary]):.3f}, AUC-PR, {np.average([x[1] for x in summary]):.3f}"
print(text, file=doc)
doc.close()
print(text)
logger.info(text)