如何从我的架构中加载模型？答案

【问题标题】：How to load the model from my architecture?如何从我的架构中加载模型？
【发布时间】：2021-03-01 07:26:19
【问题描述】：

我已经完成了这段代码，我需要加载模型以便稍后处理它，但是当我尝试使用 load_model() 时，错误是在配置文件中找不到模型。当我尝试加载权重时，错误是无法将以 HDF5 格式保存的权重加载到尚未创建其变量的子类模型中。先调用模型，然后加载权重。

这是我的代码

class Sampling(layers.Layer):
    """Uses (z_mean, z_log_var) to sample z, the vector encoding a digit."""

    def call(self, inputs):
        z_mean, z_log_var = inputs
        batch = tf.shape(z_mean)[0]
        dim = tf.shape(z_mean)[1]
        epsilon = tf.keras.backend.random_normal(shape=(batch, dim))
        return z_mean + tf.exp(0.5 * z_log_var) * epsilon

我定义了一个编码器和一个解码器，稍后会用到

class VAE(keras.Model):
    def __init__(self, encoder, decoder, **kwargs):
        super(VAE, self).__init__(**kwargs)
        self.encoder = encoder
        self.decoder = decoder

    def train_step(self, data):
        if isinstance(data, tuple):
            data = data[0]
        with tf.GradientTape() as tape:
            z_mean, z_log_var, z = encoder(data)
            reconstruction = decoder(z)
            reconstruction_loss = tf.reduce_mean(
                keras.losses.binary_crossentropy(data, reconstruction)
            )
            reconstruction_loss *= 64 * 64 * 3
            kl_loss = 1 + z_log_var - tf.square(z_mean) - tf.exp(z_log_var)
            kl_loss = tf.reduce_mean(kl_loss)
            kl_loss *= -0.5
            total_loss = reconstruction_loss + kl_loss
        grads = tape.gradient(total_loss, self.trainable_weights)
        self.optimizer.apply_gradients(zip(grads, self.trainable_weights))
        return {
            "loss": total_loss,
            "reconstruction_loss": reconstruction_loss,
            "kl_loss": kl_loss,
        }
    
    def test_step(self, data):
        if isinstance(data, tuple):
            data = data[0]
        with tf.GradientTape() as tape:
            z_mean, z_log_var, z = encoder(data)
            reconstruction = decoder(z)
            reconstruction_loss = tf.reduce_mean(
                keras.losses.binary_crossentropy(data, reconstruction)
            )
            reconstruction_loss *= 64 * 64 * 3
            kl_loss = 1 + z_log_var - tf.square(z_mean) - tf.exp(z_log_var)
            kl_loss = tf.reduce_mean(kl_loss)
            kl_loss *= -0.5
            total_loss = reconstruction_loss + kl_loss
        grads = tape.gradient(total_loss, self.trainable_weights)
        self.optimizer.apply_gradients(zip(grads, self.trainable_weights))
        return {
            "loss": total_loss,
            "reconstruction_loss": reconstruction_loss,
            "kl_loss": kl_loss,
        }

最后这就是我使用它和创建模型的方式

model_name = 'car_racing_VAE.h5'

vae = VAE(encoder, decoder)
vae.compile(optimizer=keras.optimizers.Adam(0.001))

checkpointer = keras.callbacks.ModelCheckpoint(filepath=model_name, monitor='val_loss', verbose=1, save_best_only=True, mode='min', save_freq='epoch')

history = vae.fit(train, train,
                epochs=150,
                batch_size = 128,
                shuffle=True,
                validation_data=(val, val), validation_batch_size=128,
                callbacks=[checkpointer])

那么，我怎样才能加载模型并在以后使用呢？

model = load_model(model_name)
vae.load_weights(model_name)

它们都没有工作

【问题讨论】：

那么您的编码器路径在哪里？你能附上你遇到的错误吗？
您用来保存模型的完整路径是什么？查看该位置并查看 .h5 文件是否存在。
.h5 文件是项目所在的位置。我试图创建一个文件夹并调用它，所以我尝试了 vae.load_weights(model_name) 和 vae.load_weights('/tmp/my_model.h5')。使用model = load_model('my_model.h5')时，我得到的错误是Unable to load weights saved in HDF5 format into a subclassed Model which has not created its variables yet. Call the Model first, then load the weights.或No model found in config file.

标签： tensorflow keras h5py encoder decoder

【解决方案1】：

根据错误，从路径加载保存的模型似乎存在问题。

正如 Gerry P 在评论中所建议的那样，检查保存模型的位置并从同一位置加载它。

您可以使用以下代码来避免位置问题。

#saving model
model = ...  # Get model 
model.save('path/to/location')

#loading the model back
from tensorflow import keras
model = keras.models.load_model('path/to/location')

【讨论】：