type=str,
default='./logfile',
metavar='LOG_PATH',
help='logfile path, tensorboard format')
parser.add_argument('--savedir',
type=str,
default='./models',
metavar='SAVE_PATH',
help='saving path, pickle format')
args = parser.parse_args()
args.cuda = args.cuda and torch.cuda.is_available()
np.random.seed(args.seed)
random.seed(args.seed)
torch.manual_seed(args.seed)
if args.cuda:
torch.cuda.manual_seed(args.seed)
# That is how you usually build the dataset
#dataset = CoraDataset(feature_file = './data/cora.features',
# edge_file = './data/cora_edgelist', label_file = './data/cora_label')
#dataset.read_embbedings('./embedding/embedding_line_cora')
#dataset.setting(20, 1000)
# but we load the example of cora
with open('cora.dataset', 'rb') as fdata:
dataset = pkl.load(fdata, encoding='iso-8859-1')
gan = GraphSGAN(Generator(200, dataset.k + dataset.d),
Discriminator(dataset.k + dataset.d, dataset.m), dataset,
args)
gan.train()
type=float,
default=1,
metavar='N',
help='scale factor between labeled and unlabeled data')
parser.add_argument('--logdir',
type=str,
default='./logfile',
metavar='LOG_PATH',
help='logfile path, tensorboard format')
parser.add_argument('--savedir',
type=str,
default='./models',
metavar='SAVE_PATH',
help='saving path, pickle format')
parser.add_argument('--d_repeat',
type=int,
default=5,
metavar='DR',
help='training D repeat times (default: 5)')
args = parser.parse_args()
args.cuda = args.cuda and torch.cuda.is_available()
np.random.seed(args.seed)
preprocessor = PreProcessor()
gan = ImprovedGAN(Generator(100), Discriminator(),
preprocessor.labeled_dataset(),
preprocessor.unlabeled_dataset(),
preprocessor.test_dataset(), args)
gan.train()
gan.eval()
help='random seed (default: 1)')
parser.add_argument('--log-interval', type=int, default=100, metavar='N',
help='how many batches to wait before logging training status')
parser.add_argument('--eval-interval', type=int, default=1, metavar='N',
help='how many batches to wait before evaling training status')
parser.add_argument('--unlabel-weight', type=float, default=0.5, metavar='N',
help='scale factor between labeled and unlabeled data')
parser.add_argument('--logdir', type=str, default='./logfile', metavar='LOG_PATH', help='logfile path, tensorboard format')
parser.add_argument('--savedir', type=str, default='./models', metavar='SAVE_PATH', help = 'saving path, pickle format')
args = parser.parse_args()
args.cuda = args.cuda and torch.cuda.is_available()
np.random.seed(args.seed)
random.seed(args.seed)
torch.manual_seed(args.seed)
if args.cuda:
torch.cuda.manual_seed(args.seed)
# That is how you usually build the dataset
#dataset = CoraDataset(feature_file = './data/cora.features',
# edge_file = './data/cora_edgelist', label_file = './data/cora_label')
#dataset.read_embbedings('./embedding/embedding_line_cora')
#dataset.setting(20, 1000)
# but we load the example of cora
with open('cora.dataset', 'rb') as fdata:
dataset = pkl.load(fdata, encoding='latin1')
gan = GraphSGAN(Generator(200, dataset.k + dataset.d), Discriminator(dataset.k + dataset.d, dataset.m), dataset, args)
gan.train()
'--eval-interval',
type=int,
default=1,
metavar='N',
help='how many batches to wait before evaling training status')
parser.add_argument('--unlabel-weight',
type=float,
default=1,
metavar='N',
help='scale factor between labeled and unlabeled data')
parser.add_argument('--logdir',
type=str,
default='./logfile',
metavar='LOG_PATH',
help='logfile path, tensorboard format')
parser.add_argument('--savedir',
type=str,
default='./models',
metavar='SAVE_PATH',
help='saving path, pickle format')
args = parser.parse_args()
args.cuda = args.cuda and torch.cuda.is_available()
np.random.seed(args.seed)
#gan = ImprovedGAN(Generator(100), Discriminator(), MnistLabel(10), MnistUnlabel(), MnistTest(), args)
path = "/scratch/ks4883/dl_data/"
gan = ImprovedGAN(Generator(64), Discriminator(), DL_Label(path, 60),
DL_Unlabel(path), DL_Test(path), args)
gan.train()
default='./logfile',
metavar='LOG_PATH',
help='logfile path, tensorboard format')
parser.add_argument('--savedir',
type=str,
default='./models',
metavar='SAVE_PATH',
help='saving path, pickle format')
args = parser.parse_args()
args.cuda = args.cuda and torch.cuda.is_available()
np.random.seed(args.seed)
device = torch.device("cuda:0" if args.cuda else "cpu")
cudnn.benchmark = True
gan = ImprovedGAN(Generator(100), Discriminator(), MnistLabel(10),
MnistUnlabel(), MnistVal(), args)
# gan = ImprovedGAN(Generator(100, output_dim = 64 * 64 * 3),
# Discriminator(input_dim = 64 * 64 * 3, output_dim = 1000),
# ImageNetLabel(1000, 2), ImageNetUnlabel(), ImageNetVal(),
# args)
# gan = ImprovedGAN(Generator(z_dim=100, nc=3).to(device),
# Discriminator(nc = 3, output_units = 1000).to(device),
# ImageNetLabel(1000, 2), ImageNetUnlabel(), ImageNetVal(),
# args)
gan.train()
default=1,
metavar='N',
help='how many batches to wait before evaling training status')
parser.add_argument('--unlabel-weight',
type=float,
default=1,
metavar='N',
help='scale factor between labeled and unlabeled data')
parser.add_argument('--logdir',
type=str,
default='./logfile',
metavar='LOG_PATH',
help='logfile path, tensorboard format')
parser.add_argument('--savedir',
type=str,
default='./models',
metavar='SAVE_PATH',
help='saving path, pickle format')
args = parser.parse_args()
args.cuda = args.cuda and torch.cuda.is_available()
if args.cuda:
print("Training with GPU")
np.random.seed(args.seed)
# gan = ImprovedGAN(Generator(100), Discriminator(), MnistLabel(10), MnistUnlabel(), MnistTest(), args)
gan = ImprovedGAN(Generator(z_dim=1), Discriminator(), args)
gan.train()
# gan.test = gan.data.load_train_data_sup()
# print(gan.eval() / gan.test.dataset.__len__())
# gan.test = gan.data.load_val_data()
print(gan.eval() / gan.test.dataset.__len__())
default=1,
metavar="N",
help="scale factor between labeled and unlabeled data",
)
parser.add_argument(
"--logdir",
type=str,
default="./logfile",
metavar="LOG_PATH",
help="logfile path, tensorboard format",
)
parser.add_argument(
"--savedir",
type=str,
default="./models",
metavar="SAVE_PATH",
help="saving path, pickle format",
)
args = parser.parse_args()
args.cuda = args.cuda and torch.cuda.is_available()
np.random.seed(args.seed)
gan = ImprovedGAN(
Generator(100),
Discriminator(),
MnistLabel(10),
MnistUnlabel(),
MnistTest(),
args,
)
gan.train()