#!/usr/bin/env python2.7
from argparse import ArgumentParser
parser = ArgumentParser()
parser.add_argument('--nepoch', type=int, default=50)
parser.add_argument('--version', type=str, default='4')
parser.add_argument('--trunc', type=int, default=7)
parser.add_argument('--limit', type=int, default=100)
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 = args.version
data, dims = train.instantiate(args.trunc, args.limit)
clf_gen = train.setup_data(data)
clf = train.build_classifier(dims)
train.train(clf, 'baseline', clf_gen['train'], clf_gen['validation'])
'w_clf' : 0.001,
'w_adv' : 0.1,
}
else:
opts = {
'loss' : args.adv,
'scale' : 0.1,
'w_clf' : 0.001,
'w_adv' : 1,
}
clf = train.build_classifier(dims)
if args.adv is not None:
adv = train.build_adversary(clf=clf, **opts)
preload = '%s/%s/baseline_best.h5'%(train.MODELDIR, train._APOSTLE)
if path.isfile(preload):
print 'Pre-loading weights from',preload
tmp_ = train.load_model(preload)
clf.set_weights(tmp_.get_weights())
if args.train_baseline or not(path.isfile(preload)):
train.train(clf, 'baseline', clf_gen['train'], clf_gen['validation'])
if args.adv:
print 'Training the full adversarial stack:'
callback_params = {
'partial_model' : clf,
'monitor' : lambda x : opts['w_clf'] * x.get('val_y_hat_loss') - opts['w_adv'] * x.get('val_adv_loss'), # semi-arbitrary
}
train.train(adv, args.adv, adv_gen['train'], adv_gen['validation'], callback_params)
#!/usr/bin/env python2.7
from subtlenet.models import particles as train
from argparse import ArgumentParser
parser = ArgumentParser()
parser.add_argument('--nepoch', type=int, default=50)
parser.add_argument('--version', type=int, default=4)
parser.add_argument('--trunc', type=int, default=4)
parser.add_argument('--limit', type=int, default=50)
args = parser.parse_args()
train.NEPOCH = args.nepoch
train.VERSION = args.version
data, dims = train.instantiate(args.trunc, args.limit)
clf_gen = train.setup_data(data)
clf = train.build_classifier(dims)
train.train(clf, 'classifier', clf_gen['train'], clf_gen['validation'])