学习于：简单粗暴 TensorFlow 2

1. 自定义模型

• 重载 call() 方法，pytorch 是重载 forward() 方法

import tensorflow as tf
X = tf.constant([[1.0, 2.0, 3.0],[4.0, 5.0, 6.0]])
y = tf.constant([[10.0],[20.0]])class Linear(tf.keras.Model):def __init__(self):super().__init__()self.dense = tf.keras.layers.Dense(units=1,activation=None,kernel_initializer=tf.zeros_initializer(),bias_initializer=tf.zeros_initializer())def call(self, input): # 重载 call 方法output = self.dense(input)return outputmodel = Linear()# 优化器
optimizer = tf.keras.optimizers.SGD(learning_rate=0.001)for i in range(100):with tf.GradientTape() as tape: # 梯度记录器y_pred = model(X)loss = tf.reduce_mean(tf.square(y_pred-y)) # 损失grads = tape.gradient(loss, model.variables) # 求导# 更新参数optimizer.apply_gradients(grads_and_vars=zip(grads,model.variables))

2. 学习流程

• 加载手写数字数据集

class MNistLoader():def __init__(self):data = tf.keras.datasets.mnist# 加载数据(self.train_data, self.train_label),(self.test_data, self.test_label) = data.load_data()# 扩展维度，灰度图1通道 [batch_size, 28, 28, chanels=1]self.train_data = np.expand_dims(self.train_data.astype(np.float32) / 255.0, axis=-1)self.test_data = np.expand_dims(self.test_data.astype(np.float32) / 255.0, axis=-1)self.train_label = self.train_label.astype(np.int32)self.test_label = self.test_label.astype(np.int32)# 样本个数self.num_train_data, self.num_test_data = self.train_data.shape[0], self.test_data.shape[0]def get_batch(self, batch_size):# 从训练集里随机取出 batch_size 个样本idx = np.random.randint(0, self.num_train_data, batch_size)return self.train_data[idx, :], self.train_label[idx]

• 定义模型

# 自定义多层感知机模型
class MLPmodel(tf.keras.Model):def __init__(self):super().__init__()# 除第一维以外的维度展平self.flatten = tf.keras.layers.Flatten()self.dense1 = tf.keras.layers.Dense(units=100, activation='relu')self.dense2 = tf.keras.layers.Dense(units=10)def call(self, input):x = self.flatten(input)x = self.dense1(x)x = self.dense2(x)output = tf.nn.softmax(x)return output

• 训练

# 参数
num_epochs = 5
batch_size = 50
learning_rate = 1e-4# 模型实例
mymodel = MLPmodel()
# 数据加载
data_loader = MNistLoader()
# adam 优化器
optimizer = tf.keras.optimizers.Adam(learning_rate=learning_rate)num_batches = int(data_loader.num_train_data//batch_size * num_epochs)
# 训练
for idx in range(num_batches):# 取出数据X,y = data_loader.get_batch(batch_size)with tf.GradientTape() as tape: # 梯度记录y_pred = mymodel(X) # 预测# 计算交叉熵损失loss = tf.keras.losses.sparse_categorical_crossentropy(y_true=y, y_pred=y_pred)loss = tf.reduce_mean(loss)print("batch {}, loss {}".format(idx, loss.numpy()))# 计算梯度grads = tape.gradient(loss, mymodel.variables)# 更新参数optimizer.apply_gradients(grads_and_vars=zip(grads, mymodel.variables))

• 预测

# 评估标准
sparse_categorical_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()
num_batches = int(data_loader.num_test_data // batch_size)
# 预测
for idx in range(num_batches):# 数据区间start, end = idx*batch_size, (idx+1)*batch_size# 放入模型，预测y_pred = mymodel.predict(data_loader.test_data[start : end])# 统计更新 预测信息    sparse_categorical_accuracy.update_state(y_true=data_loader.test_label[start:end],y_pred=y_pred)
print("test 准确率：{}".format(sparse_categorical_accuracy.result()))
# test 准确率：0.9455000162124634