1.Tesorflow训练模型的数据加载
- 将tensorflow的训练数据数组（矩阵）保存为.npy的数据格式。为后续的模型训练提供便捷的方法。例如如下：
import numpy as np
x=np.random.rand(100,7,9)#x是训练数据，这有100条数据，每一条有7*9个特征
np.save(r"C:\结果\y_train_feature.npy",feature)#feature是训练数据矩阵- 加载.npy训练数据和测试数组（矩阵），加载后需要调整数据的形状以满足设计模型的输入输出需求，不然无法训练模型。
import numpy as np
'''加载训练和测试数据'''
y_train_feature=np.load('C:\结果\y_train_feature.npy')
y_test_feature=np.load('C:\结果\y_test_feature.npy')
'''加载训练和测试的标签'''
y_train=np.load('C:\结果\y_train.npy')
y_test=np.load('C:\结果\y_test.npy')
'''调整数据为模型输入输出所需要的形状'''
y_test_feature=y_test_feature.reshape(-1,7,129)
y_train_feature=y_train_feature.reshape(-1,7,129)
y_test=y_test.reshape(-1,129)
y_train=y_train.reshape(-1,129)
2.Tensorflow模型搭建的训练和保存
- 这里可以采用自定义层和tensorflow的API搭建网络模型，以API为例子搭建多层LSTM网络模型用于训练。
import tensorflow as tf
from tensorflow import keras
from keras.models import Model
from keras.layers import Input, MaxPooling1D,Dense, Conv1D, Conv2D, Dropout, Flatten,\
BatchNormalization, Reshape, Activation, concatenate,SeparableConv2D,LSTM,GRU,Conv2DTranspose,SimpleRNN
from keras.layers.pooling import MaxPooling2D
'''模型输入'''
input_feature = Input(shape=y_train_feature.shape[1:], name='input_feature')
'''模型搭建LSTM'''
fc1=LSTM(512,return_sequences=True)(input_feature)
fc2=LSTM(512,return_sequences=True)(fc1)
fc3=LSTM(512,return_sequences=True)(fc2)
fc4=LSTM(512)(fc3)
'''全连接的输出层'''
output_2 = Dense(129,activation='linear',kernel_initializer='random_uniform',\
bias_initializer='random_uniform',name='output1')(fc4)
"""以Model来组合整个网络"""
model = Model(inputs= input_feature, outputs=output_2)
model.summary()#察看网络架构- tensorflow模型的训练，优化器的选择，损失函数的设置，批量大小，学习率大小，训练epoch次数，最优模型的保存等操作。
adam=Adam(lr=0.0001,decay=1e-6)#优化器的选择
model.compile(optimizer=adam,
loss='mae',#均方误差
metrics=['mae'])#平均绝对误差
'''训练网络'''
earlyStopping = callbacks.EarlyStopping(monitor='val_mae',patience=5,mode='min')
#保存模型的路径
filepath = "weights-improvement-{epoch:02d}-{val_mae:.2f}.h5"
#在每个训练期（epoch）后保存模型
#period：检查点之间的间隔（epoch数）
checkpoint = callbacks.ModelCheckpoint(filepath, monitor='val_mae', verbose=1, save_best_only=True,
mode='min',save_weights_only=False,period=1)
print('Training------')
model.fit(y_train_feature,y_train,epochs=800,batch_size=128,verbose=1,validation_data=(y_test_feature,y_test),shuffle=True,callbacks=[earlyStopping,checkpoint])
model.save(r'new_models_ripleinputV7.h5')
3.Tensorflow模型的调用
- 训练好并保存后的模型，调用时可以采用 load_model() 方法获得保存后的模型。
from keras.models import load_model
import numpy as np
model = tf.keras.models.load_model(r"保存模型的路径\LSTM.h5")
'''输入数据获得模型输出结果'''
output=model.predict(input_feature)#input_feature形状需要满足模型输入的形状要求
'''将张量形式的输出转化为numpy数组的形式'''
output=np.array(output)4.整体代码实现
- tensorflow网络模型的训练完整实现（API网络搭建）的形式。自定义的网络模型可采用同样的方法训练模型。
# -*- coding: utf-8 -*-
"""
Created on Sat Sep 11 19:47:21 2021
@author: 茶墨先生
"""
'''导入需要的包'''
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers,Sequential,datasets
from keras import regularizers,callbacks
tf.compat.v1.disable_eager_execution()
from keras import backend as K
from keras.datasets import mnist
from keras.optimizers import Adam
import numpy as np
from keras.models import Model
from keras.layers import Input, MaxPooling1D,Dense, Conv1D, Conv2D, Dropout, Flatten,\
BatchNormalization, Reshape, Activation, concatenate,SeparableConv2D,LSTM,GRU,Conv2DTranspose,SimpleRNN
from keras.layers.pooling import MaxPooling2D
#自定义模型
'''训练'''
def Train():
'''加载训练数据'''
y_train_feature=np.load('C:\结果\y_train_speech_feature.npy')
y_test_feature=np.load('C:\结果\y_test_speech_feature.npy')
y_train=np.load('C:\结果\y_train.npy')
y_test=np.load('C:\结果\y_test.npy')
y_test_feature=y_test_feature.reshape(-1,7,129)
y_train_feature=y_train_feature.reshape(-1,7,129)
y_test=y_test.reshape(-1,129)
y_train=y_train.reshape(-1,129)
print("y_train: ",y_train.shape)
print("y_train_speech_feature: ", y_train_feature.shape)
'''构建需要的网络模型'''
input_feature = Input(shape=y_train_feature.shape[1:], name='input_feature')
fc1=LSTM(512,return_sequences=True)(input_feature)
fc2=LSTM(512,return_sequences=True)(fc1)
fc3=LSTM(512,return_sequences=True)(fc2)
fc4=LSTM(512)(fc3)
output_2 = Dense(129,activation='linear',kernel_initializer='random_uniform',\
bias_initializer='random_uniform',name='output1')(fc4)
"""以Model来组合整个网络"""
model = Model(inputs= input_feature, outputs=output_2)
model.summary()
adam=Adam(lr=0.0001,decay=1e-6)
model.compile(optimizer=adam,
loss='mae',#均方误差
metrics=['mae'])#平均绝对误差
'''训练网络'''
earlyStopping = callbacks.EarlyStopping(monitor='val_mae',patience=5,mode='min')
#保存模型的路径
filepath = "weights-improvement-{epoch:02d}-{val_mae:.2f}.h5"
#在每个训练期（epoch）后保存模型
#period：检查点之间的间隔（epoch数）
checkpoint = callbacks.ModelCheckpoint(filepath, monitor='val_mae', verbose=1, save_best_only=True,
mode='min',save_weights_only=False,period=1)
print('Training------')
model.fit(y_train_feature,y_train,epochs=800,batch_size=128,verbose=1,validation_data=(y_test_feature,y_test),shuffle=True,callbacks=[earlyStopping,checkpoint])
model.save(r'new_models_ripleinputV7.h5')
'''主函数调用'''
if __name__ =='__main__':
'''训练'''
Train()
读书，生活，旅行。感谢关注和支持！文章来源地址https://www.toymoban.com/news/detail-742708.html
文章来源：https://www.toymoban.com/news/detail-742708.html
到了这里，关于Tensorflow实现训练数据的加载—模型搭建训练保存—模型调用和加载全流程的文章就介绍完了。如果您还想了解更多内容，请在右上角搜索TOY模板网以前的文章或继续浏览下面的相关文章，希望大家以后多多支持TOY模板网！
