-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathschedule.py
executable file
·157 lines (121 loc) · 5.57 KB
/
schedule.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
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
#!/usr/bin/env python3
import os
import sys
import argparse
import logging
import json
from save_to_csv import save_results
import utils.multiprocessing
from defaults import get_cfg_defaults
logger = logging.getLogger("logger")
logger.setLevel(logging.DEBUG)
ch = logging.StreamHandler(stream=sys.stdout)
ch.setLevel(logging.DEBUG)
formatter = logging.Formatter("%(asctime)s %(name)s %(levelname)s: %(message)s")
ch.setFormatter(formatter)
logger.addHandler(ch)
def f(setting):
import train_AAE
import novelty_detector
global idx
idx = 0
import torch
torch.set_default_tensor_type('torch.cuda.FloatTensor')
torch.cuda.set_device(idx)
device0 = torch.cuda.current_device()
device1 = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
#CUDA_VISIBLE_DEVICES= 5,6,7
#device = [5,6]
print("Running on GPU: %d, %s" % (idx, torch.cuda.get_device_name(device0)))
print("Running on GPU: %d, %s" % (idx, torch.cuda.get_device_name(device1)))
fold_id = setting['fold']
inliner_classes = setting['digit']
logger.debug('Using fold_id: %d', fold_id)
logger.debug('Using inliner_classes: %s', inliner_classes)
logger.debug('Percentage used: %d', cfg.DATASET.PERCENTAGES)
train_AAE.train(fold_id, [inliner_classes], inliner_classes, cfg, setting )
res = novelty_detector.main(fold_id, [inliner_classes], inliner_classes, classes_count, mul, cfg=cfg)
return res
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Scheduler for Deep Learning Model')
parser.add_argument('--min_nsample', metavar='N', type=int, help='Minimal size of Sample (mandatory)', required=True)
parser.add_argument('--max_nsample', metavar='N', type=int, help='Maximal size of Sample (mandatory)', required=True)
parser.add_argument('--batch_size', metavar='N', type=int, help='Batch Size (optional)')
parser.add_argument('--recon_scale1', metavar='X', type=float, help='Reconstruction scale factor 1 (default=2.5)', default=2.5)
parser.add_argument('--recon_scale2', metavar='X', type=float, help='Reconstruction scale factor 2 (default=2.5)', default=2.5)
parser.add_argument('--lambda_val', metavar='X', type=float, help='Lambda value (default=0.01)', default=0.01)
parser.add_argument('--percentage', metavar='X', type=int, help='Percentage (default=50)', default=50)
parser.add_argument('--full_run', metavar='V', type=bool, help='True and will run 5-fold', default=False)
parser.add_argument('--input_folder', metavar='PATH', type=str, help='Basename for Input Folder (optional)', default='INPUT')
parser.add_argument('--output_folder', metavar='PATH', type=str, help='Basename for Output Folder (optional)', default='OUTPUT' )
parser.add_argument('--config_file', metavar='PATH', type=str, help="Configure File (default='configs/mnist.yaml')", default='configs/mnist.yaml')
args = parser.parse_args()
#if len(sys.argv) > 1:
# cfg_file = 'configs/' + sys.argv[1]
#else:
# cfg_file = 'configs/mnist.yaml'
cfg_file=args.config_file
mul = 0.2
settings = []
classes_count = 10
for fold in range(5 if args.full_run else 1):
for i in range(classes_count):
settings.append(dict(fold=fold, digit=i))
cfg = get_cfg_defaults()
cfg.merge_from_file(cfg_file)
#cfg.freeze()
if args.batch_size is None:
args.batch_size=cfg.TRAIN.BATCH_SIZE
else:
cfg.TRAIN.BATCH_SIZE=args.batch_size
if args.percentage is None:
args.percentage=cfg.DATASET.PERCENTAGES
else:
cfg.DATASET.PERCENTAGES=args.percentage
if args.input_folder is None:
args.input_folder=cfg.DATASET.PATH
else:
cfg.DATASET.PATH=args.input_folder
if args.output_folder is None:
args.output_folder=cfg.OUTPUT_FOLDER
else:
cfg.OUTPUT_FOLDER=args.output_folder
cfg.freeze()
logger.debug("Min Size of Sample: %d", args.min_nsample)
logger.debug("Max Size of Sample: %d", args.max_nsample)
logger.debug("Reconstruction scale factor 1: %f", args.min_nsample)
logger.debug("Reconstruction scale factor 2: %f", args.max_nsample)
logger.debug("Lambda value: %f", args.min_nsample)
logger.debug("Batch Size : %d", args.batch_size)
logger.debug("Output folder : %s", args.output_folder)
logger.debug('Percentage : %d', args.percentage)
logger.debug("Output folder : %s", args.output_folder)
logger.debug('Config file : %s', cfg_file)
print("CONFIGURATION FILE")
print(cfg)
print()
print("SETTINGS")
print(settings)
print()
#sys.exit(1)
#gpu_count = utils.multiprocessing.get_gpu_count()
#gpu_count = min(utils.multiprocessing.get_gpu_count(), 4)
gpu_count = 1
#CUDA_VISIBLE_DEVICES=4,5,6,7
#results = utils.multiprocessing.map(f, gpu_count, settings)
results=[]
for iset in settings:
iset['min_nsample']=args.min_nsample
iset['max_nsample']=args.max_nsample
iset['batch_size']=args.batch_size
iset['recon_scale1']=args.recon_scale1
iset['recon_scale2']=args.recon_scale2
iset['lambda_val']=args.lambda_val
iset['percentage']=args.percentage
if not os.path.isdir(cfg.OUTPUT_FOLDER):
os.mkdir(cfg.OUTPUT_FOLDER)
wf=open(cfg.OUTPUT_FOLDER+os.sep+'settings.json','w')
json.dump(iset, wf, sort_keys=True, indent=4)
wf.close()
results.append( f(iset) )
save_results(results, os.path.join(cfg.OUTPUT_FOLDER, cfg.RESULTS_NAME+'_'+str(iset['fold'])+'_'+str(iset['digit'])+'.csv'))