def train_vgg11():
num_outputs = 10
architecture = ((1, 64), (1, 128), (2, 256), (2, 512), (2, 512))
net = nn.Sequential()
with net.name_scope():
net.add(vgg_stack(architecture), nn.Flatten(),
nn.Dense(4096, activation="relu"), nn.Dropout(.5),
nn.Dense(4096, activation="relu"), nn.Dropout(.5),
nn.Dense(num_outputs))
# 初始化
ctx = gpu_or_cpu()
net.initialize(ctx=ctx, init=init.Xavier())
print('initialize weight on', ctx)
# 获取数据
batch_size = 64
resize = 96
train_data, test_data = load_data(DataType.FASHION_MNIST,
batch_size,
resize=resize)
# 训练
loss = gluon.loss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(), 'sgd',
{'learning_rate': 0.05})
train(train_data, test_data, net, loss, trainer, ctx, num_epochs=20)
def train_lenet():
net = nn.Sequential()
with net.name_scope():
net.add(
nn.Conv2D(channels=20, kernel_size=5, activation='relu'),
nn.MaxPool2D(pool_size=2, strides=2),
nn.Conv2D(channels=50, kernel_size=3, activation='relu'),
nn.MaxPool2D(pool_size=2, strides=2),
nn.Flatten(),
nn.Dense(128, activation="relu"),
nn.Dense(10)
)
# 初始化
ctx = gpu_or_cpu()
net.initialize(ctx=ctx)
print('initialize weight on', ctx)
# 获取数据
batch_size = 256
train_data, test_data = load_data(DataType.FASHION_MNIST, batch_size)
# 训练
loss = gluon.loss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(), 'sgd', {'learning_rate': 0.5})
train(train_data, test_data, net, loss, trainer, ctx, num_epochs=5)
def train_nin():
net = nn.Sequential()
with net.name_scope():
net.add(
mlpconv(96, 11, 0, strides=4),
mlpconv(256, 5, 2),
mlpconv(384, 3, 1),
nn.Dropout(.5),
mlpconv(10, 3, 1, max_pooling=False),
nn.GlobalAvgPool2D(),
nn.Flatten()
)
# 初始化
ctx = gpu_or_cpu()
net.initialize(ctx=ctx, init=init.Xavier())
print('initialize weight on', ctx)
# 获取数据
batch_size = 64
resize = 224
train_data, test_data = load_data(DataType.FASHION_MNIST, batch_size, resize=resize)
# 训练
loss = gluon.loss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(), 'sgd', {'learning_rate': 0.1})
train(train_data, test_data, net, loss, trainer, ctx, num_epochs=20)
def train_alexnet():
net = nn.Sequential()
with net.name_scope():
net.add(
# 第一阶段
nn.Conv2D(channels=96,
kernel_size=11,
strides=4,
activation='relu'),
nn.MaxPool2D(pool_size=3, strides=2),
# 第二阶段
nn.Conv2D(channels=256,
kernel_size=5,
padding=2,
activation='relu'),
nn.MaxPool2D(pool_size=3, strides=2),
# 第三阶段
nn.Conv2D(channels=384,
kernel_size=3,
padding=1,
activation='relu'),
nn.Conv2D(channels=384,
kernel_size=3,
padding=1,
activation='relu'),
nn.Conv2D(channels=256,
kernel_size=3,
padding=1,
activation='relu'),
nn.MaxPool2D(pool_size=3, strides=2),
# 第四阶段
nn.Flatten(),
nn.Dense(4096, activation="relu"),
nn.Dropout(.5),
# 第五阶段
nn.Dense(4096, activation="relu"),
nn.Dropout(.5),
# 第六阶段
nn.Dense(10))
# 初始化
ctx = gpu_or_cpu()
net.initialize(ctx=ctx, init=init.Xavier())
print('initialize weight on', ctx)
# 获取数据
batch_size = 64
resize = 224
train_data, test_data = load_data(DataType.FASHION_MNIST,
batch_size,
resize=resize)
# 训练
loss = gluon.loss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(), 'sgd',
{'learning_rate': 0.01})
train(train_data, test_data, net, loss, trainer, ctx, num_epochs=20)
def train_resnet():
# 初始化
ctx = gpu_or_cpu()
net = ResNet(10)
net.initialize(ctx=ctx, init=init.Xavier())
print('initialize weight on', ctx)
# 获取数据
batch_size = 64
resize = 96
train_data, test_data = load_data(DataType.FASHION_MNIST,
batch_size,
resize=resize)
# 训练
loss = gluon.loss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(), 'sgd',
{'learning_rate': 0.05})
train(train_data, test_data, net, loss, trainer, ctx, num_epochs=20)
train_ds = vision.ImageFolderDataset(os.path.join(input_str, 'train'),
flag=1)
valid_ds = vision.ImageFolderDataset(os.path.join(input_str, 'valid'),
flag=1)
test_ds = vision.ImageFolderDataset(os.path.join(input_str, 'test'),
flag=1)
loader = gluon.data.DataLoader
train_valid_data = loader(train_valid_ds.transform_first(transform_train),
batch_size,
shuffle=True,
last_batch='keep')
train_data = loader(train_ds.transform_first(transform_train),
batch_size,
shuffle=True,
last_batch='keep')
valid_data = loader(valid_ds.transform_first(transform_test),
batch_size,
shuffle=True,
last_batch='keep')
test_data = loader(test_ds.transform_first(transform_test),
batch_size,
shuffle=False,
last_batch='keep')
ctx = gpu_or_cpu()
num_outputs = 10
filename = "data/mlp.params"
# train_cifar10(ctx, num_outputs, filename, train_data, valid_data)
classification(ctx, num_outputs, filename, test_data)