parser.add_argument('--lr', type=float, default=0.1)
parser.add_argument('--momentum', type=float, default=0.0)
parser.add_argument('--weightdecay', type=float, default=0.0)
parser.add_argument('--model', type=str, default='vgg')
parser.add_argument('--resume', type=str, default=None)
parser.add_argument('--datadir',
type=str,
default='/Users/wjf/datasets/SVHN/train25k_test70k')
parser.add_argument('--logdir', type=str, default='logs/SGDCov')
args = parser.parse_args()
logger = LogSaver(args.logdir)
logger.save(str(args), 'args')
# data
dataset = SVHN(args.datadir)
logger.save(str(dataset), 'dataset')
test_list = dataset.getTestList(1000, True)
# model
start_iter = 0
lr = args.lr
if args.model == 'resnet':
from resnet import ResNet18
model = ResNet18().cuda()
elif args.model == 'vgg':
from vgg import vgg11
model = vgg11().cuda()
else:
raise NotImplementedError()
criterion = CEwithMask
parser.add_argument('--maxiter', type=int, default=int(2e5 + 1))
parser.add_argument('--lr', type=float, default=0.05)
parser.add_argument('--list-size', type=int, default=5000)
parser.add_argument('--sigma', type=float, default=1e-3)
parser.add_argument('--resume', type=str, default=None)
parser.add_argument('--datadir',
type=str,
default='/home/wjf/datasets/SVHN/train25000_test70000')
parser.add_argument('--logdir', type=str, default='logs/GLD')
args = parser.parse_args()
logger = LogSaver(args.logdir)
logger.save(str(args), 'args')
# data
dataset = SVHN(args.datadir)
logger.save(str(dataset), 'dataset')
train_list = dataset.getTrainList(args.list_size, True)
test_list = dataset.getTestList(1000, True)
# model
start_iter = 0
model = vgg11().cuda()
logger.save(str(model), 'classifier')
criterion = nn.CrossEntropyLoss().cuda()
optimizer = torch.optim.SGD(model.parameters(), lr=args.lr)
logger.save(str(optimizer), 'optimizer')
if args.resume:
checkpoint = torch.load(args.resume)
start_iter = checkpoint['iter']
print "Force restoring..."
ckpt_file = FLAGS.restore_path
# data_directory = os.path.join(FLAGS.data_dir, "mnist")
else:
# assert False
data_directory = FLAGS.data_dir
ckpt_file = os.path.join(data_directory, "svhn" + param_string + ".ckpt")
# data_directory = os.path.join(FLAGS.data_dir, "mnist")
if not os.path.exists(data_directory):
os.makedirs(data_directory)
print "Using ckpt file:", ckpt_file
# train_data = mnist.input_data.read_data_sets(mnist_directory, one_hot=True).train # binarized (0-1) mnist train data
# test_data = mnist.input_data.read_data_sets(mnist_directory, one_hot=True).test # binarized (0-1) mnist test data
svhn_data = SVHN()
train_data = svhn_data.train()
test_data = svhn_data.test()
fetches = []
fetches.extend([Lx, Lz, train_op])
Lxs = [0] * train_iters
Lzs = [0] * train_iters
sess = tf.InteractiveSession()
saver = tf.train.Saver() # saves variables learned during training
tf.initialize_all_variables().run()
#saver.restore(sess, "/tmp/draw/drawmodel.ckpt") # to restore from model, uncomment this line
# e_values = sess.run(my_e,feed_dict={})
'As preprocessing; scale the image randomly between 2 numbers and crop randomly at networs input size'
)
tf.app.flags.DEFINE_string('train_root_dir', '../training',
'Root directory to put the training data')
tf.app.flags.DEFINE_integer('log_step', 10000,
'Logging period in terms of iteration')
NUM_CLASSES = 10
TRAIN_FILE = 'svhn'
TEST_FILE = 'mnist'
print TRAIN_FILE + ' ---------------------------------------> ' + TEST_FILE
print TRAIN_FILE + ' ---------------------------------------> ' + TEST_FILE
print TRAIN_FILE + ' ---------------------------------------> ' + TEST_FILE
TRAIN = SVHN('data/svhn', split='train', shuffle=True)
VALID = MNIST('data/mnist', split='test', shuffle=True)
TEST = MNIST('data/mnist', split='test', shuffle=False)
FLAGS = tf.app.flags.FLAGS
MAX_STEP = 10000
def decay(start_rate, epoch, num_epochs):
return start_rate / pow(1 + 0.001 * epoch, 0.75)
def adaptation_factor(x):
#return 1.0
#return 0.25
den = 1.0 + math.exp(-10 * x)
def bias_variable(shape):
return tf.Variable(tf.constant(1.0, shape=shape))
def convolution(x, w):
return tf.nn.conv2d(x, w, strides=[1, 1, 1, 1], padding="SAME")
def max_pool(x):
return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
# Load data
svhn = SVHN("../res", n_classes, use_extra=True, gray=False)
# Create the model
X = tf.placeholder(tf.float32, [None, image_size, image_size, channels])
Y = tf.placeholder(tf.float32, [None, n_classes])
# Weights & Biases
weights = {
"layer1": weight_variable([filter_size, filter_size, channels, depth_1]),
"layer2": weight_variable([filter_size, filter_size, depth_1, depth_2]),
"layer3": weight_variable([filter_size, filter_size, depth_2, depth_3]),
"layer4": weight_variable([image_size // 8 * image_size // 8 * depth_3, hidden]),
"layer5": weight_variable([hidden, n_classes])
}
parser = argparse.ArgumentParser()
parser.add_argument('--maxiter', type=int, default=int(2e5+1))
parser.add_argument('--lr', type=float, default=0.05)
parser.add_argument('--batch-size', type=int, default=100)
parser.add_argument('--list-size', type=int, default=5000)
parser.add_argument('--update-noise', type=int, default=10)
parser.add_argument('--resume', type=str, default=None)
parser.add_argument('--datadir', type=str, default='/home/wjf/datasets/SVHN/train25000_test70000')
parser.add_argument('--logdir', type=str, default='logs/GLD_diag')
args = parser.parse_args()
logger = LogSaver(args.logdir)
logger.save(str(args), 'args')
# data
dataset = SVHN(args.datadir)
logger.save(str(dataset), 'dataset')
train_list = dataset.getTrainList(args.list_size, True)
test_list = dataset.getTestList(1000, True)
# model
start_iter = 0
model = vgg11().cuda()
logger.save(str(model), 'classifier')
criterion = nn.CrossEntropyLoss().cuda()
optimizer = torch.optim.SGD(model.parameters(), lr=args.lr)
logger.save(str(optimizer), 'optimizer')
if args.resume:
checkpoint = torch.load(args.resume)
start_iter = checkpoint['iter']
def main(args):
args.cuda = not args.no_cuda and torch.cuda.is_available()
init_seeds(seed=int(time.time()))
kwargs = {'num_workers': 2, 'pin_memory': True} if args.cuda else {}
print(args.dataset)
if args.dataset == 'MNIST':
test_dataloader = data.DataLoader(
MNIST(args.data_path, args.run_folder, transform=mnist_transformer()),
batch_size=10000, shuffle=False, **kwargs)
train_dataset = MNIST(args.data_path, args.run_folder, train=True, transform=mnist_transformer(), imbalance_ratio=args.imbalance_ratio)
if args.imbalance_ratio == 100:
args.num_images = 25711
else:
args.num_images = 50000
args.budget = 125
args.initial_budget = 125
args.num_classes = 10
args.num_channels = 1
args.arch_scaler = 2
elif args.dataset == 'SVHN':
test_dataloader = data.DataLoader(
SVHN(args.data_path, args.run_folder, transform=svhn_transformer()),
batch_size=5000, shuffle=False, **kwargs)
train_dataset = SVHN(args.data_path, args.run_folder, train=True, transform=svhn_transformer(), imbalance_ratio=args.imbalance_ratio)
if args.imbalance_ratio == 100:
args.num_images = 318556
else:
args.num_images = 500000
args.budget = 1250
args.initial_budget = 1250
args.num_classes = 10
args.num_channels = 3
args.arch_scaler = 1
elif args.dataset == 'cifar10':
test_dataloader = data.DataLoader(
datasets.CIFAR10(args.data_path, download=True, transform=cifar_transformer(), train=False),
batch_size=args.batch_size, drop_last=False)
train_dataset = CIFAR10(args.data_path)
args.num_images = 50000
args.budget = 2500
args.initial_budget = 5000
args.num_classes = 10
args.num_channels = 3
elif args.dataset == 'cifar100':
test_dataloader = data.DataLoader(
datasets.CIFAR100(args.data_path, download=True, transform=cifar_transformer(), train=False),
batch_size=args.batch_size, drop_last=False)
train_dataset = CIFAR100(args.data_path)
args.num_images = 50000
args.budget = 2500
args.initial_budget = 5000
args.num_classes = 100
args.num_channels = 3
elif args.dataset == 'ImageNet':
test_dataloader = data.DataLoader(
ImageNet(args.data_path + '/val', transform=imagenet_test_transformer()),
batch_size=args.batch_size, shuffle=False, drop_last=False, **kwargs)
if args.imbalance_ratio == 100:
train_dataset = ImageNet(args.data_path + '/train_ir_100', transform=imagenet_train_transformer())
args.num_images = 645770
else:
train_dataset = ImageNet(args.data_path + '/train', transform=imagenet_train_transformer())
args.num_images = 1281167
args.budget = 64000
args.initial_budget = 64000
args.num_classes = 1000
args.num_channels = 3
args.arch_scaler = 1
else:
raise NotImplementedError
all_indices = set(np.arange(args.num_images))
initial_indices = random.sample(all_indices, args.initial_budget)
sampler = data.sampler.SubsetRandomSampler(initial_indices)
#print(args.batch_size, sampler)
# dataset with labels available
querry_dataloader = data.DataLoader(train_dataset, sampler=sampler,
batch_size=args.batch_size, drop_last=False, **kwargs)
print('Sampler size =', len(querry_dataloader))
solver = Solver(args, test_dataloader)
splits = range(1,11)
current_indices = list(initial_indices)
accuracies = []
for split in splits:
print("Split =", split)
# need to retrain all the models on the new images
# re initialize and retrain the models
#task_model = vgg.vgg16_bn(num_classes=args.num_classes)
if args.dataset == 'MNIST':
task_model = model.LeNet(num_classes=args.num_classes)
elif args.dataset == 'SVHN':
task_model = resnet.resnet10(num_classes=args.num_classes)
elif args.dataset == 'ImageNet':
task_model = resnet.resnet18(num_classes=args.num_classes)
else:
print('WRONG DATASET!')
# loading pretrained
if args.pretrained:
print("Loading pretrained model", args.pretrained)
checkpoint = torch.load(args.pretrained)
task_model.load_state_dict({k: v for k, v in checkpoint['state_dict'].items() if 'fc' not in k}, strict=False) # copy all but last linear layers
#
vae = model.VAE(z_dim=args.latent_dim, nc=args.num_channels, s=args.arch_scaler)
discriminator = model.Discriminator(z_dim=args.latent_dim, s=args.arch_scaler)
#print("Sampling starts")
unlabeled_indices = np.setdiff1d(list(all_indices), current_indices)
unlabeled_sampler = data.sampler.SubsetRandomSampler(unlabeled_indices)
unlabeled_dataloader = data.DataLoader(train_dataset, sampler=unlabeled_sampler,
batch_size=args.batch_size, drop_last=False, **kwargs)
#print("Train starts")
# train the models on the current data
acc, vae, discriminator = solver.train(querry_dataloader,
task_model,
vae,
discriminator,
unlabeled_dataloader)
print('Final accuracy with {}% of data is: {:.2f}'.format(int(split*100.0*args.budget/args.num_images), acc))
accuracies.append(acc)
sampled_indices = solver.sample_for_labeling(vae, discriminator, unlabeled_dataloader)
current_indices = list(current_indices) + list(sampled_indices)
sampler = data.sampler.SubsetRandomSampler(current_indices)
querry_dataloader = data.DataLoader(train_dataset, sampler=sampler,
batch_size=args.batch_size, drop_last=False, **kwargs)
torch.save(accuracies, os.path.join(args.out_path, args.log_name))
def bias_variable(shape):
return tf.Variable(tf.constant(1.0, shape=shape))
def conv2d(x, w):
return tf.nn.conv2d(x, w, strides=[1, 1, 1, 1], padding="SAME")
def max_pool(x):
return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
# 读取数据集
svhn = SVHN(10, use_extra=False, gray=False)
# 建立模型
X = tf.placeholder(tf.float32, [None, 32, 32, 3])
Y = tf.placeholder(tf.float32, [None, 10])
# Batch normalization
mean, variance = tf.nn.moments(X, [1, 2, 3], keep_dims=True)
x = tf.nn.batch_normalization(X, mean, variance, normalization_offset, normalization_scale,normalization_epsilon)
#第一层卷积
W_conv1 = weight_variable([5, 5, 3, 16])
nb_classes = 10
nb_epoch = 20
path = '/home/...' # 修改路径
# 输入图像的维度，SVHN库的图片大小均为32*32
img_rows, img_cols = 32, 32
# 卷积层中使用的卷积核的个数
nb_filters = 32
# 池化层操作的范围
pool_size = (2, 2)
# 卷积核的大小
kernel_size = (3, 3)
# 加载SVHN数据库中的图片，并设置训练集和测试集
svhn = SVHN('/home/cv503/Desktop/conv_visualization/svhn_dataset',
nb_classes,
gray=False)
X_train = svhn.train_data
Y_train = svhn.train_labels
X_test = svhn.test_data
Y_test = svhn.test_labels
# 后端使用tensorflow时，即tf模式下，
# 第一个维度是样本维，表示样本的数目，
# 第二和第三个维度是高和宽，
# 最后一个维度是通道维，表示颜色通道数
X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 3)
X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 3)
input_shape = (img_rows, img_cols, 3)
# 将X_train, X_test的数据格式转为float32
parser.add_argument('--ghostsize', type=int, default=100)
parser.add_argument('--numghost', type=int, default=100)
parser.add_argument('--update-noise', type=int, default=10)
parser.add_argument('--momentum', type=float, default=0.0)
parser.add_argument('--weightdecay', type=float, default=0.0)
parser.add_argument('--model', type=str, default='vgg')
parser.add_argument('--resume', type=str, default=None)
parser.add_argument('--datadir', type=str, default='/Users/wjf/datasets/SVHN/train25k_test70k')
parser.add_argument('--logdir', type=str, default='logs/GLD_Fisher')
args = parser.parse_args()
logger = LogSaver(args.logdir)
logger.save(str(args), 'args')
# data
dataset = SVHN(args.datadir)
logger.save(str(dataset), 'dataset')
train_list = dataset.getTrainList(args.batchsize, True)
test_list = dataset.getTestList(1000, True)
# model
start_iter = 0
lr = args.lr
if args.model == 'resnet':
from resnet import ResNet18
model = ResNet18().cuda()
elif args.model == 'vgg':
from vgg import vgg11
model = vgg11().cuda()
else:
raise NotImplementedError()
def main():
global args, best_prec1
args = parser.parse_args()
# Use specific GPU
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu
devices = [int(s) for s in args.gpu.split(',') if s.isdigit()]
nGPU = len(devices)
devices = list(range(nGPU))
args.distributed = args.world_size > 1
if args.distributed:
dist.init_process_group(backend=args.dist_backend,
init_method=args.dist_url,
world_size=args.world_size)
# create model
if args.pretrained:
print("=> using pre-trained model '{}'".format(args.arch))
model = models.__dict__[args.arch](pretrained=True, r=args.se_reduce)
else:
print("=> creating model '{}'".format(args.arch))
# model = models.__dict__[args.arch](r=args.se_reduce)
model = models.__dict__[args.arch]()
if not args.distributed:
if args.arch.startswith('alexnet') or args.arch.startswith('vgg'):
model.features = torch.nn.DataParallel(model.features)
model.cuda()
else:
if (nGPU == 1):
model.cuda()
else:
model = torch.nn.DataParallel(model, device_ids=devices).cuda()
else:
model.cuda()
model = torch.nn.parallel.DistributedDataParallel(model)
print(args)
print(model)
# define loss function (criterion) and optimizer
criterion = nn.CrossEntropyLoss().cuda()
# optimizer = torch.optim.SGD(model.parameters(),
# args.lr,
# momentum=args.momentum,
# weight_decay=args.weight_decay)
lr_policy = list()
params_collect = dict(model.named_parameters())
for k, v in params_collect.items():
if 'st' in k:
lr_policy.append({
'params': v,
'lr': 0.001,
'weight_decay': args.weight_decay
})
else:
lr_policy.append({'params': v})
optimizer = torch.optim.SGD(lr_policy,
args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
# optionally resume from a checkpoint
if args.resume:
if os.path.isfile(args.resume):
print("=> loading checkpoint '{}'".format(args.resume))
checkpoint = torch.load(args.resume)
args.start_epoch = checkpoint['epoch']
best_prec1 = checkpoint['best_prec1']
model.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
print("=> loaded checkpoint '{}' (epoch {})".format(
args.resume, checkpoint['epoch']))
else:
print("=> no checkpoint found at '{}'".format(args.resume))
cudnn.benchmark = True
# Data loading code
if (args.dataset == 'imagenet'):
traindir = os.path.join(args.data, 'train')
valdir = os.path.join(args.data, 'val')
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
train_dataset = datasets.ImageFolder(
traindir,
transforms.Compose([
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize,
]))
if args.distributed:
train_sampler = torch.utils.data.distributed.DistributedSampler(
train_dataset)
else:
train_sampler = None
train_loader = torch.utils.data.DataLoader(
train_dataset,
batch_size=args.batch_size,
shuffle=(train_sampler is None),
num_workers=args.workers,
pin_memory=True,
sampler=train_sampler)
val_loader = torch.utils.data.DataLoader(datasets.ImageFolder(
valdir,
transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
normalize,
])),
batch_size=args.batch_size,
shuffle=False,
num_workers=args.workers,
pin_memory=True)
elif (args.dataset == 'cifar10'):
to_normalized_tensor = [
transforms.ToTensor(),
# transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))]
transforms.Normalize((0.4914, 0.4822, 0.4465),
(0.2470, 0.2435, 0.2616))
]
data_augmentation = [
transforms.RandomCrop(28, padding=0),
transforms.RandomHorizontalFlip()
]
transform = transforms.Compose(data_augmentation +
to_normalized_tensor)
trainset = datasets.CIFAR10(root='./data',
train=True,
download=True,
transform=transform)
train_loader = torch.utils.data.DataLoader(trainset,
batch_size=args.batch_size,
shuffle=True,
num_workers=2)
testset = datasets.CIFAR10(root='./data',
train=False,
download=True,
transform=transform)
val_loader = torch.utils.data.DataLoader(testset,
batch_size=args.batch_size,
shuffle=False,
num_workers=2)
elif (args.dataset == 'svhn'):
path_to_train_lmdb_dir = os.path.join(args.data, 'train.lmdb')
path_to_val_lmdb_dir = os.path.join(args.data, 'val.lmdb')
train_loader = torch.utils.data.DataLoader(
SVHN(path_to_train_lmdb_dir),
batch_size=args.batch_size,
shuffle=True,
num_workers=2,
pin_memory=True)
val_loader = torch.utils.data.DataLoader(SVHN(path_to_val_lmdb_dir),
batch_size=128,
shuffle=False)
# to_normalized_tensor = [transforms.ToTensor(),
# # transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))]
# transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2470, 0.2435, 0.2616))]
# data_augmentation = [transforms.RandomCrop(28, padding=0),
# transforms.RandomHorizontalFlip()]
#
# transform = transforms.Compose(data_augmentation + to_normalized_tensor)
#
# trainset = datasets.CIFAR10(root='./data', train=True,
# download=True, transform=transform)
# train_loader = torch.utils.data.DataLoader(trainset, batch_size=args.batch_size,
# shuffle=True, num_workers=2)
#
# testset = datasets.CIFAR10(root='./data', train=False,
# download=True, transform=transform)
# val_loader = torch.utils.data.DataLoader(testset, batch_size=args.batch_size,
# shuffle=False, num_workers=2)
else:
print('No dataset named a ', args.dataset)
exit(0)
if args.evaluate:
validate(val_loader, model, criterion)
print(acc)
return
print("Number of parameters: ",
sum(p.numel() for p in model.parameters() if p.requires_grad))
for epoch in range(args.start_epoch, args.epochs):
if args.distributed:
train_sampler.set_epoch(epoch)
adjust_learning_rate(optimizer, epoch)
print("Learning Rate: ", optimizer.param_groups[0]['lr'])
# train for one epoch
train(train_loader, model, criterion, optimizer, epoch)
# evaluate on validation set
prec1 = validate(val_loader, model, criterion)
# remember best prec@1 and save checkpoint
is_best = prec1 > best_prec1
best_prec1 = max(prec1, best_prec1)
save_checkpoint(
{
'epoch': epoch + 1,
'arch': args.arch,
'state_dict': model.state_dict(),
'best_prec1': best_prec1,
'optimizer': optimizer.state_dict(),
}, is_best, args.save + '/checkpoint.pth.tar')
print("\nBest Model: ", best_prec1)
def __init__(self,
base_dataset='svhn',
take_amount=None,
take_amount_seed=13,
add_svhn_extra=True,
aux_data_filename=None,
add_aux_labels=False,
aux_take_amount=None,
train=False,
**kwargs):
"""A dataset with auxiliary pseudo-labeled data"""
if base_dataset == 'svhn':
if train:
self.dataset = SVHN(split='train', **kwargs)
else:
self.dataset = SVHN(split='test', **kwargs)
# elif base_dataset == 'mnist_m':
# if train:
# self.dataset = MNIST_M(split='train', **kwargs)
# else:
# self.dataset = MNIST_M(split='test', **kwargs)
# elif base_dataset == "syndigit" :
# if train:
# self.dataset = SynDigit(split="train" , **kwargs)
# else:
# self.dataset = SynDigit(split="test" , **kwargs)
else:
raise ValueError('Dataset %s not supported' % base_dataset)
# because torchvision is annoying
self.dataset.targets = self.dataset.labels
self.targets = list(self.targets)
if train and add_svhn_extra:
svhn_extra = SVHN(split='noise_20p', **kwargs)
self.data = np.concatenate([self.data, svhn_extra.data])
self.targets.extend(svhn_extra.labels)
self.base_dataset = base_dataset
self.train = train
if self.train:
if take_amount is not None:
rng_state = np.random.get_state()
np.random.seed(take_amount_seed)
take_inds = np.random.choice(len(self.sup_indices),
take_amount, replace=False)
np.random.set_state(rng_state)
logger = logging.getLogger()
logger.info('Randomly taking only %d/%d examples from training'
' set, seed=%d, indices=%s',
take_amount, len(self.sup_indices),
take_amount_seed, take_inds)
self.targets = self.targets[take_inds]
self.data = self.data[take_inds]
self.sup_indices = list(range(len(self.targets)))
self.unsup_indices = []
if aux_data_filename is not None:
aux_path = os.path.join(kwargs['root'], aux_data_filename)
print("Loading data from %s" % aux_path)
with open(aux_path, 'rb') as f:
aux = pickle.load(f)
aux_data = aux['data']
aux_targets = aux['extrapolated_targets']
orig_len = len(self.data)
if aux_take_amount is not None:
rng_state = np.random.get_state()
np.random.seed(take_amount_seed)
take_inds = np.random.choice(len(aux_data),
aux_take_amount, replace=False)
np.random.set_state(rng_state)
logger = logging.getLogger()
logger.info(
'Randomly taking only %d/%d examples from aux data'
' set, seed=%d, indices=%s',
aux_take_amount, len(aux_data),
take_amount_seed, take_inds)
aux_data = aux_data[take_inds]
aux_targets = aux_targets[take_inds]
self.data = np.concatenate((self.data, aux_data), axis=0)
if not add_aux_labels:
self.targets.extend([-1] * len(aux_data))
else:
self.targets.extend(aux_targets)
# note that we use unsup indices to track the labeled datapoints
# whose labels are "fake"
self.unsup_indices.extend(
range(orig_len, orig_len+len(aux_data)))
logger = logging.getLogger()
logger.info("--Training set--")
logger.info("Number of training samples: %d", len(self.targets))
logger.info("Number of supervised samples: %d",
len(self.sup_indices))
logger.info("Number of unsup samples: %d", len(self.unsup_indices))
logger.info("Label (and pseudo-label) histogram: %s",
tuple(
zip(*np.unique(self.targets, return_counts=True))))
logger.info("Shape of training data: %s", np.shape(self.data))
# Test set
else:
self.sup_indices = list(range(len(self.targets)))
self.unsup_indices = []
logger = logging.getLogger()
logger.info("--Test set--")
logger.info("Number of samples: %d", len(self.targets))
logger.info("Label histogram: %s",
tuple(
zip(*np.unique(self.targets, return_counts=True))))
logger.info("Shape of data: %s", np.shape(self.data))
parser = argparse.ArgumentParser()
parser.add_argument('--maxiter', type=int, default=int(1e5 + 1))
parser.add_argument('--lr', type=float, default=0.05)
parser.add_argument('--batch-size', type=int, default=100)
parser.add_argument('--resume', type=str, default=None)
parser.add_argument('--datadir',
type=str,
default='/home/wjf/datasets/SVHN/train25000_test70000')
parser.add_argument('--logdir', type=str, default='logs/SGD')
args = parser.parse_args()
logger = LogSaver(args.logdir)
logger.save(str(args), 'args')
# data
dataset = SVHN(args.datadir)
logger.save(str(dataset), 'dataset')
test_list = dataset.getTestList(1000, True)
# model
start_iter = 0
model = vgg11().cuda()
logger.save(str(model), 'classifier')
criterion = nn.CrossEntropyLoss().cuda()
optimizer = torch.optim.SGD(model.parameters(), lr=args.lr)
logger.save(str(optimizer), 'optimizer')
if args.resume:
checkpoint = torch.load(args.resume)
start_iter = checkpoint['iter']
model.load_state_dict(checkpoint['model'])
import tensorflow as tf
from svhn import SVHN
n_input = 3072 # warstwa wejściowa(32x32x3 piksele i rgb)
n_hidden1 = 512 # pierwsza warstwa ukryta
n_hidden2 = 256 # druga warstwa ukryta
n_hidden3 = 128 # trzecia warstwa ukryta
n_output = 10 # warstwa wyjściowa (cyfry od 0 do 9)
svhn = SVHN("../res", n_output, use_extra=False, gray=False)
learning_rate = 0.001
batch_size = 40
n_iterations = int(svhn.train_examples / batch_size)
normalization_offset = 0.0 # beta
normalization_scale = 1.0 # gamma
normalization_epsilon = 0.001 # epsilon
#placeholdery na wejściowe i wyjściowe dane
X = tf.placeholder("float", [None, 32, 32, 3], name="X")
Y = tf.placeholder("float", [None, n_output], name="Y")
#wagi pomiędzy warstwami
weights = {
'w1': tf.Variable(tf.truncated_normal([n_input, n_hidden1], stddev=0.1)),
'w2': tf.Variable(tf.truncated_normal([n_hidden1, n_hidden2], stddev=0.1)),
'w3': tf.Variable(tf.truncated_normal([n_hidden2, n_hidden3], stddev=0.1)),
'out': tf.Variable(tf.truncated_normal([n_hidden3, n_output], stddev=0.1)),
}
