重置 Keras model 的所有权重

Question

我希望能够重置整个 Keras model 的权重，这样我就不必再次编译它了。 编译 model 目前是我代码的主要瓶颈。 这是我的意思的一个例子：

import tensorflow as tf  

model = tf.keras.Sequential([
    tf.keras.layers.Flatten(input_shape=(28, 28)),
    tf.keras.layers.Dense(16, activation='relu'),
    tf.keras.layers.Dense(10)
])

model.compile(optimizer=tf.keras.optimizers.SGD(learning_rate=0.001),
                loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
                metrics=['accuracy'])
   
data = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = data.load_data()

model.fit(x=x_train, y=y_train, epochs=10)

# Reset all weights of model here
# model.reset_all_weights() <----- something like that

model.fit(x=x_train, y=y_train, epochs=10)

Answer 1

你可以使用这个循环：

for ix, layer in enumerate(model.layers):
    if hasattr(model.layers[ix], 'kernel_initializer') and \
            hasattr(model.layers[ix], 'bias_initializer'):
        weight_initializer = model.layers[ix].kernel_initializer
        bias_initializer = model.layers[ix].bias_initializer

        old_weights, old_biases = model.layers[ix].get_weights()

        model.layers[ix].set_weights([
            weight_initializer(shape=old_weights.shape),
            bias_initializer(shape=len(old_biases))])

原始重量：

model.layers[1].get_weights()[0][0]

array([ 0.4450057 , -0.13564804,  0.35884023,  0.41411972,  0.24866664,
        0.07641453,  0.45726687, -0.04410008,  0.33194816, -0.1965386 ,
       -0.38438258, -0.13263905, -0.23807487,  0.40130925, -0.07339832,
        0.20535922], dtype=float32)

新权重：

model.layers[1].get_weights()[0][0]

array([-0.4607593 , -0.13104361, -0.0372932 , -0.34242013,  0.12066692,
       -0.39146423,  0.3247317 ,  0.2635846 , -0.10496247, -0.40134245,
        0.19276887,  0.2652442 , -0.18802321, -0.18488845,  0.0826562 ,
       -0.23322225], dtype=float32)

Answer 2

我写了一个 function 重新初始化 tensorflow 2 中的权重。

def reinitialize(model):
    for l in model.layers:
        if hasattr(l,"kernel_initializer"):
            l.kernel.assign(l.kernel_initializer(tf.shape(l.kernel)))
        if hasattr(l,"bias_initializer"):
            l.bias.assign(l.bias_initializer(tf.shape(l.bias)))
        if hasattr(l,"recurrent_initializer"):
            l.recurrent_kernel.assign(l.recurrent_initializer(tf.shape(l.recurrent_kernel)))

我花了比它应该想出的时间更长的时间，我尝试了许多在我的特定用例中失败的事情。 IMO 这应该是标准的 TF 功能。