40、数据处理与特征预处理全解析

数据处理与特征预处理全解析

1. 自定义训练循环

若想构建自定义训练循环，可自然地对训练集进行迭代：

for X_batch, y_batch in train_set:
    [...] # 执行一次梯度下降步骤

还能创建一个执行整个训练循环的 TF 函数：

import tensorflow as tf

@tf.function
def train(model, optimizer, loss_fn, n_epochs, [...]):
    train_set = csv_reader_dataset(train_filepaths, repeat=n_epochs, [...])
    for X_batch, y_batch in train_set:
        with tf.GradientTape() as tape:
            y_pred = model(X_batch)
            main_loss = tf.reduce_mean(loss_fn(y_batch, y_pred))
            loss = tf.add_n([main_loss] + model.losses)
        grads = tape.gradient(loss, model.trainable_variables)
        optimizer.apply_gradients(zip(grads, model.trainable_variables)