parser.add_argument('--trunc',type=int,default=7)
parser.add_argument('--limit',type=int,default=100)
parser.add_argument('--adv',type=str,default=None)
parser.add_argument('--train_baseline',action='store_true')
args = parser.parse_args()
import extra_vars
from subtlenet.models import particles as train
from os import path
train.NEPOCH = args.nepoch
train.VERSION = str(args.version) + '_Adam'
#train.OPTIMIZER = 'RMSprop'
data, dims = train.instantiate(args.trunc, args.limit)
clf_gen = train.setup_data(data)
adv_gen = train.setup_adv_data(data)
if args.adv == 'emd':
opts = {
'loss' : train.emd,
'scale' : 0.1,
'w_clf' : 0.001,
'w_adv' : 100,
}
elif args.adv == 'mse':
opts = {
'loss' : args.adv,
'scale' : 0.03,
'w_clf' : 0.001,
import extra_vars
from subtlenet.models import particles as train
from os import path
from subtlenet import config
from subtlenet.backend import obj
# config.DEBUG = True
# obj._RANDOMIZE = False
train.NEPOCH = args.nepoch
train.VERSION = str(args.version) + '_Adam'
#train.OPTIMIZER = 'RMSprop'
data, dims = train.instantiate(args.trunc, args.limit)
clf_gen = train.setup_data(data)
adv_gen = train.setup_data(data, decorr_mass=True)
if args.adv == 'emd':
opts = {
'loss': train.emd,
'scale': 0.1,
'w_clf': 0.001,
'w_adv': 100,
}
elif args.adv == 'mse':
opts = {
'loss': args.adv,
'scale': 0.03,
'w_clf': 0.001,
'w_adv': 0.1,