def is_valid(self, url, target):
img = self.get_img(url)
img = self.deal_with_img(img)
model = AlexNet()
model = torch.nn.DataParallel(model)
optimizer = optim.SGD(model.parameters(), lr=LR)
scheduler = StepLR(optimizer, step_size=20, gamma=0.1)
checkpoint = torch.load('AlexNet.pkl', map_location='cpu')
if checkpoint is not None:
model.load_state_dict(checkpoint["net"])
optimizer.load_state_dict(checkpoint["optimizer"])
return self.test(model, img) == target
def __init__(self, input_shape, num_classes, learning_rate, clients_num):
# 返回一个上下文管理器，这个上下管理器使用这个图作为要操作的图
self.graph = tf.Graph()
# tf.Session()创建一个会话，当上下文管理器退出时会话关闭和资源释放自动完成。
self.sess = tf.Session(graph=self.graph)
# 调用create函数来构建AlexNet的计算图
# `net` 是一个list，依次包含模型中FedModel需要的计算节点（看上面）
net = AlexNet(input_shape, num_classes, learning_rate, self.graph)
print("net done")
self.model = FedModel(*net)
# 初始化
# Variable：作为存储节点的变量（Variable）不是一个简单的节点，而是一副由四个子节点构成的子图：
# （1）变量的初始值——initial_value。
# （2）更新变量值的操作op——Assign。
# （3）读取变量值的操作op——read
# （4）变量操作——（a）
# 上述四个步骤即：首先，将initial_value赋值（Assign）给节点，存储在（a）当中，当需要读取该变量时，调用read函数取值即可
with self.graph.as_default():
self.sess.run(tf.global_variables_initializer())
# Load Cifar-10 dataset
# NOTE: len(self.dataset.train) == clients_num
# 加载数据集。对于训练集：`self.dataset.train[56]`可以获取56号client的数据集
# `self.dataset.train[56].next_batch(32)`可以获取56号client的一个batch，大小为32
# 对于测试集，所有client共用一个测试集，因此：
# `self.dataset.test.next_batch(1000)`将获取大小为1000的数据集（无随机）
# 这里已经把数据集划分成了多份，且每个client都有各自的一份
self.dataset = Dataset(load_data_mnist, split=clients_num)
def __init__(self, input_shape, num_classes, learning_rate, clients_num):
self.graph = tf.Graph()
self.sess = tf.Session(graph=self.graph)
# Call the create function to build the computational graph of AlexNet
net = AlexNet(input_shape, num_classes, learning_rate, self.graph)
self.model = FedModel(*net)
# initialize
with self.graph.as_default():
self.sess.run(tf.global_variables_initializer())
# Load Cifar-10 dataset
# NOTE: len(self.dataset.train) == clients_num
self.dataset = Dataset(tf.keras.datasets.cifar10.load_data,
split=clients_num)
def __init__(self, input_shape, num_classes, learning_rate, clients_num):
self.graph = tf.Graph()
self.sess = tf.Session(graph=self.graph)
# 创建alxnet网络
net = AlexNet(input_shape, num_classes, learning_rate, self.graph)
#net = vgg_net(input_shape, num_classes, learning_rate, self.graph)
self.model = FedModel(*net)
# 初始化
with self.graph.as_default():
self.sess.run(tf.global_variables_initializer())
# 装载数据
# 根据训练客户端数量划分数据集
self.dataset = Dataset(tf.keras.datasets.cifar10.load_data,split=clients_num)
train_images, train_labels, y = load_pokemon()
train_labels = to_categorical(train_labels, num_classes=num_classes)
print(logger_name + ' train_images shape: {}'.format(train_images.shape))
print(logger_name +
' train_images image shape: {}'.format(train_images[0].shape))
print(logger_name + ' train_labels shape: {}'.format(train_labels.shape))
print(logger_name + ' y image shape: {}'.format(y.shape))
# Data normalization
train_images = train_images.astype('float32')
y = y.astype('float32')
# train_images /= 255
# y /= 255
model = AlexNet(train_images[0].shape, num_classes)
model.summary()
# log_dir = "logs/alextnetfit/" + datetime.datetime.now().strftime("%Y%m%d-%H%M%S")
# tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=log_dir, histogram_freq=1)
model.fit(train_images, y=train_labels, verbose=0)
prediction_values = model.predict_classes(y)
# 看一下預測結果
fig = plt.figure(figsize=(15, 7))
fig.subplots_adjust(left=0,
right=1,
bottom=0,
top=1,