-
Notifications
You must be signed in to change notification settings - Fork 1
/
main.py
94 lines (70 loc) · 2.99 KB
/
main.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
from ast import parse
from multiprocessing import cpu_count
import os
import numpy as np
import random
import torch
from argparse import ArgumentParser
from pytorch_lightning import Trainer
from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint, ModelSummary, Callback
from pytorch_lightning.core.saving import save_hparams_to_yaml
from pytorch_lightning.loggers import TensorBoardLogger
from model import MMoE
seed = 2023
if not seed:
seed = random.randint(1, 10000)
print("seed is %d" %seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
np.random.seed(seed)
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.deterministic = True
logger = TensorBoardLogger(name="logs", save_dir="./")
print("gpu is available: ", torch.cuda.is_available())
if torch.cuda.is_available(): dev = "gpu"
else: dev = "cpu"
def train_test_model(hparams, seed=None):
print("Loading Model...")
model = MMoE(hparams, seed)
print("Model Built...")
early_stop = EarlyStopping(
monitor='val_loss', patience=5, verbose=True, mode='min'
)
cpkt_callback = ModelCheckpoint(
monitor='val_loss', save_top_k=1, mode='min'
)
callback = [cpkt_callback, early_stop]
trainer = Trainer(max_epochs=int(10) , callbacks=callback, logger=logger,\
devices=1, accelerator=dev
)
trainer.fit(model, train_dataloaders=model.mydataloader(train='train'), val_dataloaders=model.mydataloader(train="train"))
print("=========Train over============")
test_result = trainer.test(model, dataloaders=model.mydataloader(train="test"))
def arg_parse_():
parser = ArgumentParser()
# dataset hparams
parser.add_argument('--dataset' ,type=str, default="SMD", choices=["SMD", "SWaT", "WADI"])
parser.add_argument('--data_name', type=str)
parser.add_argument('--window', default=16, type=int)
# n_multiv means the dimension of metrics
parser.add_argument('--n_multiv', type=int)
parser.add_argument('--horize', default=1, type=int)
parser.add_argument('--batch_size', type=int, default=128)
# model hparams
parser.add_argument('--num_experts', default=5, type=int)
parser.add_argument('--n_kernel', type=int, default=16)
parser.add_argument('--experts_out', type=int, default=128)
parser.add_argument('--experts_hidden', type=int, default=256)
parser.add_argument('--towers_hidden', type=int, default=32)
parser.add_argument('--criterion', default="l2", type=str, choices=["l1", "l2"])
parser.add_argument('--exp_dropout', type=float, default=0.2)
parser.add_argument('--tow_dropout', type=float, default=0.1)
parser.add_argument('--conv_dropout', type=float, default=0.1)
parser.add_argument('--sg_ratio', type=float, default=0.7)
parser.add_argument('--lr', type=float, default=0.001)
return parser
if __name__ == "__main__":
parser = arg_parse_()
hparams = parser.parse_args()
print(hparams)
train_test_model(hparams, seed)