neural network学习8 keras中meter的使用

import os

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics
import datetime
from matplotlib import pyplot as plt
import io

# datasets用于数据集的管理，layers用于dense层，optimizer优化器，Sequential容器，metrics测试用的度量器


# 对数据集作预处理(对每一个x，y样本)
def preprocess(x, y):
    x = tf.cast(x, dtype=tf.float32) / 255.
    y = tf.cast(y, dtype=tf.int32)
    return x, y

batchsz = 128
(x, y), (x_val, y_val) = datasets.mnist.load_data()
db = tf.data.Dataset.from_tensor_slices((x, y))
db = db.map(preprocess).shuffle(60000).batch(batchsz).repeat(10)

db_val = tf.data.Dataset.from_tensor_slices((x_val, y_val))
db_val = db_val.map(preprocess).batch(batchsz, drop_remainder=True)  # test部分不需要shuffle

# 构建一个5层的网络
model = Sequential([
    layers.Dense(256, activation=tf.nn.relu),  # [b,784]---->[b,256]
    layers.Dense(128, activation=tf.nn.relu),  # [b,256]---->[b,128]
    layers.Dense(64, activation=tf.nn.relu),  # [b,128]---->[b,64]
    layers.Dense(32, activation=tf.nn.relu),  # [b,64]---->[b,32]
    layers.Dense(10)  # [b,32]---->[b,10] 最后一层不用激活函数 330=32*10+10
])
model.build(input_shape=(None, 28 * 28))  # 完成一个创建工作
model.summary()  # 起到调试作用，打印网络结构

# 优化器更新过程 w=w-lr*dw 只需要传入一个list即可对所有的w进行更新
optimizer = optimizers.Adam(lr=0.01)  # Adam只是一种参数最优化的方法，其他的还有SGD、Momentum

#step1 build a meter
acc_meter = metrics.Accuracy()
loss_meter = metrics.Mean()

for step, (x, y) in enumerate(db):
    with tf.GradientTape() as tape:
        x = tf.reshape(x, (-1, 28 * 28))
        out = model(x)
        y_onehot = tf.one_hot(y, depth=10)
        loss = tf.reduce_mean(tf.losses.categorical_crossentropy(y_onehot, out, from_logits=True))
        loss_meter.update_state(loss)#跟新

    grads = tape.gradient(loss, model.trainable_variables)  # 求所有导数
    optimizer.apply_gradients(zip(grads, model.trainable_variables))  # 参数原地更新

    if step % 100 == 0:
        print(step, 'loss:', loss_meter.result().numpy())# 打印出来并清零
        loss_meter.reset_states()#并清零



    if step % 500 == 0:
        total_correct = 0.
        total_num = 0.
        acc_meter.reset_states()#清零
        for _,(x, y) in enumerate(db_val):
            x = tf.reshape(x, (-1, 28 * 28))
            # test的时候不需要求梯度
            out = model(x)
            # logits=>prob
            pred = tf.argmax(out, axis=1)  # 得到最大值的索引 int64
            pred = tf.cast(pred,dtype=tf.int32)

            correct = tf.reduce_sum(tf.cast(tf.equal(pred, y), dtype=tf.int32)).numpy()
            total_correct += correct  # correct类型是tensor，total_correct类型是numpy
            total_num += x.shape[0]  # 加上batch的个数
            acc_meter.update_state(y,pred)#数据送进去

        acc = total_correct / total_num
        print(step, 'test acc:', total_correct/total_num,acc_meter.result().numpy())#输出