03、tf-keras与keras的相同和不同及实战案例 - 东方耀AI技术分享

# -*- coding: utf-8 -*-
__author__ = u'东方耀微信：dfy_88888'
__date__ = '2019/10/28 下午3:37'
__product__ = 'PyCharm'
__filename__ = 'tf_keras_classify_model'
import os
import sys
import time
import tensorflow as tf
import matplotlib as mpl
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import sklearn
from tensorflow import keras
# tf的keras 和单独的keras
# pip3 install keras
import keras
print(tf.__version__)
print(sys.version_info)
for module in mpl, np, pd, sklearn, tf, keras:
print(module.__name__, module.__version__)
# https://tensorflow.google.cn/api_docs/python/tf/keras/datasets/fashion_mnist/load_data
fashion_mnist = keras.datasets.fashion_mnist
(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()
print(X_train.shape, y_train.shape)
print('测试数据集：', X_test.shape, y_test.shape)
X_valid, X_train = X_train[:5000], X_train[5000:]
y_valid, y_train = y_train[:5000], y_train[5000:]
print('验证数据集：', X_valid.shape, y_valid.shape)
print('训练数据集：', X_train.shape, y_train.shape)
def show_single_image(img_arr):
plt.imshow(img_arr, cmap='binary')
plt.show()
# show_single_image(X_train[11])
def show_multi_images(n_rows, n_cols, X_data, y_data, class_names):
assert len(X_data) == len(y_data), '样本的特征与标签长度一致'
assert n_rows*n_cols <= len(X_data)
# 使用plt的子图 width, height in inches
plt.figure(figsize=(n_cols*1.4, n_rows*1.6))
for i in range(n_rows):
for j in range(n_cols):
index = i*n_cols + j
plt.subplot(n_rows, n_cols, index+1)
plt.imshow(X_data[index], cmap='binary', interpolation='nearest')
plt.axis('off')
plt.title(label=class_names[y_data[index]])
plt.show()
class_names = ['t-shirt', 'trouser', 'pullover', 'dress',
'coat', 'sandal', 'shirt', 'sneaker',
'bag', 'ankle boot']
# show_multi_images(3, 5, X_train[:20], y_train[:20], class_names)
# tf.keras.Sequential 构建模型结构
# model = keras.Sequential(name='tf2.0_model')
# model.add(keras.layers.Flatten(input_shape=(28, 28)))
# model.add(keras.layers.Dense(units=300, activation='relu'))
# model.add(keras.layers.Dense(units=100, activation='relu'))
# model.add(keras.layers.Dense(10, activation='softmax'))
model = keras.Sequential(layers=[
keras.layers.Flatten(input_shape=(28, 28)),
keras.layers.Dense(units=300, activation='relu'),
keras.layers.Dense(units=100, activation='relu'),
keras.layers.Dense(10, activation='softmax')
], name='tf2.0_model_2')
model.compile(optimizer=keras.optimizers.Adam(learning_rate=0.001),
loss='sparse_categorical_crossentropy',
metrics=['accuracy', 'mse'])
model.summary()
# print(tf.test.is_gpu_available())
# 开始训练
# validation_freq 是根据epochs来的每隔多少轮进行验证
history = model.fit(x=X_train, y=y_train, epochs=10, verbose=1,
validation_data=(X_valid, y_valid), validation_freq=1)
print(type(history))
# 模型训练过程中的历史数据指标
print(history.history)
# validation_freq=2 的情况：
# {'loss': [2.276887303126942, 0.5237461835861206, 0.45976711897850037, 0.4327715507962487],
# 'accuracy': [0.75996363, 0.8186182, 0.8376727, 0.8458727],
# 'mse': [27.68526, 27.679394, 27.680891, 27.681673],
# 'val_loss': [0.5112990405797958, 0.45343393814563754],
# 'val_accuracy': [0.823199987411499, 0.843999981880188],
# 'val_mse': [27.655563354492188, 27.657318115234375]}
def plot_learning_curve(history):
# ValueError: arrays must all be same length
# 表格型数据要求每一列的len一致这里即：history.history字典里每个key对应的value长度一致
df_history = pd.DataFrame(data=history.history)
print(df_history)
print(df_history.index)
print(df_history.columns)
print(df_history.dtypes)
df_history.plot(figsize=(8, 5))
plt.grid(True)
# x就是DataFrame的索引
plt.ylim(0, 1.3)
plt.show()
plot_learning_curve(history)

复制代码

from sklearn.preprocessing import StandardScaler, MinMaxScaler
# 数据的归一化处理
std_scaler = StandardScaler()
# print('训练数据集0：', X_train[0])
X_train_std = std_scaler.fit_transform(X_train.reshape(-1, 28*28))
X_train_std = X_train_std.reshape(-1, 28, 28)
print(X_train_std.shape)
X_valid_std = std_scaler.transform(X_valid.reshape(-1, 28*28))
X_valid_std = X_valid_std.reshape(-1, 28, 28)
print(X_valid_std.shape)
X_test_std = std_scaler.transform(X_test.reshape(-1, 28*28))
X_test_std = X_test_std.reshape(-1, 28, 28)

复制代码

# 测试模型
test_loss, test_accuracy, test_mse = model.evaluate(X_test_std, y_test)
print(test_loss, test_accuracy, test_mse)
y_test_predict = model.predict(X_test_std, verbose=1)
print(y_test_predict.shape)
print(y_test_predict[:2])

复制代码

# Callbacks: utilities called at certain points during model training.
# 实战回调函数
logdir = './callbacks'
if not os.path.exists(logdir):
os.mkdir(logdir)
output_model_file = os.path.join(logdir, 'fashion_mnist_model.h5')
callbacks = [
keras.callbacks.TensorBoard(logdir),
keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=1e-3, patience=5),
keras.callbacks.ModelCheckpoint(output_model_file, save_best_only=True)
]
history = model.fit(x=X_train_std, y=y_train, epochs=10, verbose=1,
validation_data=(X_valid_std, y_valid), validation_freq=1, callbacks=callbacks)
print(type(history))
# 模型训练过程中的历史数据指标
print(history.history)

复制代码