def __init__(self, device=None, learning_rate=1e-3, act=F.relu, n_cls=10, batch_size=64, n_samples=50000):
# Settings
self.device = device
self.act = act
self.learning_rate = learning_rate
self.n_cls = n_cls
self.batch_size = batch_size
self.iters = 0
self.iter_epoch = int(n_samples / batch_size)
self.epoch = 0
self.basedir = "./{}".format(int(time.time()))
if not os.path.exists(self.basedir):
os.makedirs(self.basedir)
# Loss
self.recon_loss = ReconstructionLoss()
# Model
from st.cifar10.cnn_model_003 import Model
self.model = Model(device, act)
self.model.to_gpu(device) if device is not None else None
# Optimizer
self.optimizer = optimizers.Adam(learning_rate)
self.optimizer.setup(self.model)
self.optimizer.use_cleargrads()
def __init__(self, device=None, learning_rate=1e-3, act=F.relu, n_cls=10):
super(Experiment003, self).__init__(
device=device,
learning_rate=learning_rate,
act=act,
n_cls=n_cls
)
# Loss
self.recon_loss = ReconstructionLoss()
self.er_loss = EntropyRegularizationLoss()
def __init__(self, device=None, learning_rate=1e-3, act=F.relu, n_cls=10):
# Settings
self.device = device
self.act = act
self.learning_rate = learning_rate
self.n_cls = n_cls
# Loss
self.recon_loss = ReconstructionLoss()
# Model
from st.mnist.cnn_model_001 import Model
self.model = Model(device, act)
self.model.to_gpu(device) if device is not None else None
# Optimizer
self.optimizer = optimizers.Adam(learning_rate)
self.optimizer.setup(self.model)
self.optimizer.use_cleargrads()