本文共 1352 字，大约阅读时间需要 4 分钟。

三种创建model方式

import numpy as npimport tensorflow as tffrom tensorflow import kerasrows = 10000columns = 100emb_size = 5words_length = 50000train_x = np.random.random(size=(rows, columns, emb_size))train_y = np.random.randint(low=0, high=2, size=(rows, 1))

1

model = keras.Sequential(name="test1")model.add(keras.layers.Input(shape=(columns, emb_size)))model.add(keras.layers.SimpleRNN(units=10))model.add(keras.layers.Dense(1))model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])model.fit(train_x, train_y, epochs=10, batch_size=100)print("-------------------------------------------------")model = Sequential(	[keras.layers.Input(shape=(columns, emb_size)),	keras.layers.SimpleRNN(units=10),	keras.layers.Dense(1)	])model.compile(loss="mse", optimizer="sgd")model.fit(train_x, train_y)

2

x = keras.layers.Input(shape=(columns, emb_size))y = keras.layers.SimpleRNN(units=10)(x)y = keras.layers.Dense(1)(y)model = keras.Model(inputs=x, outputs=y)model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])model.fit(train_x, train_y, epochs=10, batch_size=100)

3

class MyModel(keras.layer.Model):	def __init__(self):		pass	def call(self, input):		passmodel = MyModel()model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])model.fit()

为什么需要三种创建模式，三者有什么区别

从上往下一种比一种更接近底层，可以任意调整参数

转载地址：http://blugf.baihongyu.com/