|
|
@ -0,0 +1,55 @@ |
|
|
|
import os |
|
|
|
import tensorflow as tf |
|
|
|
import matplotlib.pyplot as plt |
|
|
|
import numpy as np |
|
|
|
|
|
|
|
mnist = tf.keras.datasets.mnist |
|
|
|
(x_train, y_train),(x_test, y_test) = mnist.load_data(path="/data/data/mnist.npz") #加载mnist数据集 |
|
|
|
|
|
|
|
#验证mnist数据集大小。x为数据,y为标签。mnist每张图的像素为28*28 |
|
|
|
print(x_train.shape) |
|
|
|
print(y_train.shape) |
|
|
|
print(x_test.shape) |
|
|
|
print(y_test.shape) |
|
|
|
|
|
|
|
#打印训练集中前9张,看看是什么数字 |
|
|
|
for i in range(9): |
|
|
|
plt.subplot(3,3,1+i) |
|
|
|
plt.imshow(x_train[i], cmap='gray') |
|
|
|
plt.show() |
|
|
|
|
|
|
|
#打印相应的标签 |
|
|
|
print(y_train[:9]) |
|
|
|
|
|
|
|
#基操:将像素标准化一下 |
|
|
|
x_train, x_test = x_train / 255.0, x_test / 255.0 |
|
|
|
|
|
|
|
#搭建一个两层神经网络 |
|
|
|
model = tf.keras.models.Sequential([ |
|
|
|
tf.keras.layers.Flatten(input_shape=(28, 28)), #拉伸图像成一维向量 |
|
|
|
tf.keras.layers.Dense(128, activation='relu'), #第一层全连接+ReLU激活 |
|
|
|
tf.keras.layers.Dropout(0.2), #dropout层 |
|
|
|
tf.keras.layers.Dense(10, activation='softmax') #第二层全连接+softmax激活,输出预测标签 |
|
|
|
]) |
|
|
|
|
|
|
|
#设置训练超参,优化器为sgd,损失函数为交叉熵,训练衡量指标为accuracy |
|
|
|
model.compile(optimizer='sgd', loss='sparse_categorical_crossentropy', metrics=['accuracy']) |
|
|
|
|
|
|
|
#开始训练,训练5个epoch,一个epoch代表所有图像计算一遍。每一个epoch能观察到训练精度的提升 |
|
|
|
model.fit(x_train, y_train, epochs=5) |
|
|
|
|
|
|
|
#计算训练了5个epoch的模型在测试集上的表现 |
|
|
|
model.evaluate(x_test, y_test) |
|
|
|
|
|
|
|
|
|
|
|
#直观看一下模型预测结果,打印测试集中的前9张图像 |
|
|
|
for i in range(9): |
|
|
|
plt.subplot(3,3,1+i) |
|
|
|
plt.imshow(x_test[i], cmap='gray') |
|
|
|
plt.show() |
|
|
|
|
|
|
|
#打印模型识别的数字,是否正确? |
|
|
|
np.argmax(model(x_test[:9]).numpy(), axis=1) |
|
|
|
|
|
|
|
#保存训练好的模型 |
|
|
|
model.save("/data/output/model_epoch_5") |