【Tensorflow+keras】使用keras API保存模型权重、plot画loss损失函数、保存训练loss值

淡淡的烟草味﹌ 2022-10-06 15:53 189阅读 0赞

# 举例实现 #

（1）模型实现

import tensorflow  as tf
    from tensorflow.keras.layers import *
    from tensorflow.keras import *
    import json
    import numpy
    # 这个类解决json.dump(dict)时报错Object of type 'float32' is not JSON serializable
    class NumpyEncoder(json.JSONEncoder):  
        def default(self, obj):  
            if isinstance(obj, (numpy.int_, numpy.intc, numpy.intp, numpy.int8,  
                numpy.int16, numpy.int32, numpy.int64, numpy.uint8,  
                numpy.uint16, numpy.uint32, numpy.uint64)):  
                return int(obj)  
            elif isinstance(obj, (numpy.float_, numpy.float16, numpy.float32,numpy.float64)):  
                return float(obj)  
            elif isinstance(obj, (numpy.ndarray,)):  
                return obj.tolist()  
            return json.JSONEncoder.default(self, obj)  
    def main()
    	# 搭建模型
    	inputs = tf.keras.layers.Input(shape=(3,))
    	d = tf.keras.layers.Dense(2, name='out')
    	output_1 = d(inputs)
    	output_2 = d(inputs)
    	model = tf.keras.models.Model(
    	inputs=inputs, outputs=[output_1, output_2])
    	model.compile(optimizer="Adam", loss="mse", metrics=["mae", "acc"])
    	# 保存模型权重
    	checkpoint = callbacks.ModelCheckpoint('real_weight_10.tf',save_format='tf', monitor='val_acc',verbose=0, save_best_only=True, mode='min', save_weights_only=True)
    	history = model.fit(x, (y, y)))
    	# 画loss曲线
    	epochs=range(len(history['bit_err']))
    	plt.figure()
    	plt.plot(epochs,history['bit_err'],'b',label='Training bit_error')
    	plt.plot(epochs,history['val_bit_err'],'r',label='Validation bit_error')
    	plt.title('Traing and Validation bit_error')
    	plt.legend()
    	plt.savefig('figure/model_bit_err_SNR10.jpg')
    	plot.show()
    	plt.figure()
    	plt.plot(epochs,history['loss'],'b',label='Training loss')
    	plt.plot(epochs,history['val_loss'],'r',label='Validation val_loss')
    	plt.title('Traing and Validation loss')
    	plt.legend()
    	plt.savefig('figure/model_loss_SNR10.jpg')
    	plt.show()
    	# 保存loss值
        history_dict = history.history
        json.dump(history_dict, open('model_history/history.json', 'w'),cls=NumpyEncoder)
        
    if __name__ == '__main__':
       # freeze_support() here if program needs to be frozen
        main()

（2）单独加载模型loss值

import numpy as np 
    
    import scipy.io as sio
    import matplotlib.pyplot as plt
    import json
    
    history = json.load(open('model_history/history.json', 'r'))
    epochs=range(len(history['bit_err']))
    plt.figure()
    plt.plot(epochs,history['bit_err'],'b',label='Training bit_error')
    plt.plot(epochs,history['val_bit_err'],'r',label='Validation bit_error')
    plt.title('Traing and Validation bit_error')
    plt.legend()
    # plt.savefig('figure/model_bit_err_SNR10.jpg')
    plot.show()
    
    plt.figure()
    plt.plot(epochs,history['loss'],'b',label='Training loss')
    plt.plot(epochs,history['val_loss'],'r',label='Validation val_loss')
    plt.title('Traing and Validation loss')
    plt.legend()
    # plt.savefig('figure/model_loss_SNR10.jpg')
    plt.show()