def main():
args = arguments.args
args.update(params)
args['loss_func'] = F.cross_entropy
torch.manual_seed(args['seed'])
np.random.seed(args['seed'])
model = net.MLP(params['x_dim'], params['width'], params['y_dim'])
# model = net.Linear(params['x_dim'], params['y_dim'])
train_data = ToyColoredMNIST(train=True)
valid_data = ToyColoredMNIST(train=True)
test_data = ToyColoredMNIST(train=False)
train_loader = DataLoader(train_data,
batch_size=params['batch_size'],
shuffle=True)
valid_loader = DataLoader(valid_data,
batch_size=params['batch_size'],
shuffle=True)
test_loader = DataLoader(test_data,
batch_size=params['test_batch_size'],
shuffle=True)
model = trainer(model, train_loader, valid_loader, test_loader, args)
new_data = ToyColoredMNIST(train=True)
get_loss_var_across_colors(args, model, new_data)
return
def __init__(self, modelName='20160818_MNIST.model'):
self.modelName = modelName
self.window = self.create_window()
# set canvas
self.image1 = Image.new("RGB", (window_width, window_height),
(255, 255, 255))
self.draw = ImageDraw.Draw(self.image1)
# set neural network model
self.mlp = net.MLP(784, 1000, 10)
model = L.Classifier(self.mlp)
serializers.load_hdf5(self.modelName, model)
def main():
args = arguments.args
args.update(params)
args['loss_func'] = F.mse_loss
model = net.MLP(params['x_dim'], params['width'], params['y_dim'])
convex_data = ConvexDataset()
train_loader = DataLoader(convex_data,
batch_size=params['batch_size'],
shuffle=False)
trainer(model, train_loader, args)
return
def main(args):
if args.model == 'cnn':
net = net_module.CNN()
else:
net = net_module.MLP(28 * 28, 10, 100)
gpu_device = args.gpu
if gpu_device >= 0:
chainer.cuda.get_device(gpu_device).use()
net.to_gpu(gpu_device)
xp = cuda.cupy
else:
xp = np
serializers.load_npz(args.model_file, net)
image = Image.open(args.image_file).convert('L').resize((28, 28),
Image.BILINEAR)
# 学習データは値の範囲が0～1なのでそれに合わせるために255で割る
# 学習データは背景が0なので反転する
image = 1 - xp.asarray(image).astype(np.float32) / 255
image = image.reshape((1, -1))
probs = cuda.to_cpu(predict(net, image))[0]
results = sorted(zip(six.moves.range(10), probs), key=lambda x: -x[1])
for n, p in results:
print('{0:d}: {1:.4f}'.format(n, p))
type=int,
default=1,
metavar='S',
help='random seed (default: 1)')
parser.add_argument('--coeff-type',
default='all-one',
choices=('all-one', 'linear', 'harmonic'),
help='coefficent function used to weight neurons')
parser.add_argument('--output',
default='models/model.pth',
help='output directory')
args = parser.parse_args()
args.cuda = torch.cuda.is_available()
torch.manual_seed(args.seed)
if args.cuda:
torch.cuda.manual_seed(args.seed)
data = datasets.get_dataset(args.dataset_root, args.dataset,
args.batch_size, args.cuda)
train_dataset, train_loader, test_dataset, test_loader = data
x, _ = train_loader.__iter__().next()
B, C, W, H = x.shape
model = net.MLP(C * W * H, 10, coeff_type=args.coeff_type)
if args.cuda:
model = model.cuda()
ps = filter(lambda x: x.requires_grad, model.parameters())
train_model(model, args.output, train_loader, test_loader, args.lr,
args.epochs)
parser.add_argument('--gpu', '-g', type=int, default=-1, help='GPU device index, -1 indicates CPU')
parser.add_argument('--epoch', '-e', type=int, default=100, help='Number of epochs')
parser.add_argument('--batch-size', '-b', type=int, default=100, help='Mini batch size')
parser.add_argument('--prefix', '-p', type=str, default=None, help='prefix of saved file name')
args = parser.parse_args()
n_epoch = args.epoch
batch_size = args.batch_size
if args.prefix is None:
prefix = args.model
else:
prefix = args.prefix
if args.model == 'cnn':
net = net_module.CNN()
else:
net = net_module.MLP(28 * 28, 10, 100)
gpu_device = args.gpu
if gpu_device >= 0:
chainer.cuda.get_device(gpu_device).use()
net.to_gpu(gpu_device)
xp = cuda.cupy
else:
xp = np
optimizer = optimizers.Adam()
optimizer.setup(net)
# MNISTデータセットを読み込む
# get_mnistはMNISTデータセットファイルがなければダウンロードを行うので
# 初回実行時は時間がかかる
# データセットは"~/.chainer/dataset"以下に保存される
train_data, test_data = chainer.datasets.get_mnist()
def __init__(self, action_space, state_dim, gamma=1):
self.logits_net = net.MLP(input_dim=state_dim,
output_dim=action_space.__len__())
self.optim = Adam(self.logits_net.parameters(), lr=0.01)
self.action_space = action_space
else:
assert False, "Invalid choice for rejection sampling"
torch.manual_seed(args.seed)
if args.cuda:
torch.cuda.manual_seed(args.seed)
train_set, train_loader, test_set, test_loader = datasets.get_randomref_dataset("",
args.batch_size,
{'train' : args.train_size,
'test' : args.test_size},
args.numreferencenodes,
args.quantization,
args.rejectionsampling,
args.noise_scale,
args.ref_scale)
adhoc_model = net.MLP(2*args.numreferencenodes, 1, activation=True)
models = {'adhoc_model' : adhoc_model}
if not os.path.exists(args.output):
os.makedirs(args.output)
out_str = "&".join(simulate_args_from_namespace(args, positional=['output']))
out_str = args.output+out_str+".pth"
print(out_str)
if args.cuda:
for _, model in models.items():
model = model.cuda()
model_acc = train_model(models, out_str, train_loader, test_loader, args.lr, args.epochs,
sisr_thres=args.sisr_thres)
