Keras模型保存的几个方法和它们的区别

model.save()

model_save_path = "model_file_path.h5"
# 保存模型
model.save(model_save_path)
# 删除当前已存在的模型
del model
# 加载模型
from keras.models import load_model
model = load_model(model_save_path)

model.save_weights()

model_save_path = "model_file_path.h5"
# 保存模型权重
model.save_weights(model_save_path)
# 加载模型权重
model.load_weights(model_save_path)

model.to_json()

model_save_path = "model_file_path.h5"
# 保存模型权重
model.save_weights(model_save_path)
# 加载模型权重
model.load_weights(model_save_path)

model.to_yaml()

# 保存模型网络结构
yaml_string = model.to_yaml()
with open("model_save_file.yaml", "w") as f:
	f.write(yaml_string)  # 将模型转为yaml文件后的字符串写入本地
# 读取模型网络结构	
from keras.models import model_from_yaml
with open("model_save_file.yaml", "r") as f:
	yaml_string = f.read()  # 读取本地模型的yaml文件
model = model_from_yaml(yaml_string)  # 创建一个模型

四种保存模型的联系与区别

转自: https://blog.csdn.net/weixin_38353277/article/details/120151402