资源 localhost/total/N10tensorflow3VarE 不存在

Question

我正在使用 Google Colab 并尝试使用 VGG 块训练 model。 像这样：

METRICS = [
      keras.metrics.TruePositives(name='tp'),
      keras.metrics.FalsePositives(name='fp'),
      keras.metrics.TrueNegatives(name='tn'),
      keras.metrics.FalseNegatives(name='fn'), 
      keras.metrics.BinaryAccuracy(name='accuracy'),
      keras.metrics.Precision(name='precision'),
      keras.metrics.Recall(name='recall'),
      keras.metrics.AUC(name='auc'),
]

# function for creating a vgg block
def vgg_block(layer_in, n_filters, n_conv):
  # add convolutional layers
  for _ in range(n_conv):
    layer_in = Conv2D(n_filters, (3,3), padding='same', activation='relu')(layer_in)
  # add max pooling layer
  layer_in = MaxPooling2D((2,2), strides=(2,2))(layer_in)
  return layer_in

# define model input
visible = Input(shape=(256, 256, 3))
# add vgg module
layer = vgg_block(visible, 64, 2)

#####################################


flat = Flatten()(layer)
hidden1 = Dense(128, activation='relu')(flat)
output = Dense(1, activation='sigmoid')(hidden1)
model = Model(inputs=visible, outputs=output)
print(model.summary())

# plot model architecture
plot_model(model, show_shapes=True, to_file='vgg_block.png')

model.compile(loss='binary_crossentropy',
            optimizer='rmsprop',
            metrics=METRICS)

# New lines to obtain the best model in term of validation accuracy
from keras.callbacks import ModelCheckpoint
filepath="weights-improvement-{epoch:02d}-{val_accuracy:.2f}.h5"
checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
callbacks_list = [checkpoint]

但是，当我尝试使用 model.fit_generator 时，它给了我一个错误。 我正在使用的代码是：

history = model.fit_generator(
    train_generator,
    steps_per_epoch=2000 // batch_size,
    epochs=20,
    validation_data=validation_generator,
    validation_steps=800 // batch_size,
    callbacks=callbacks_list
)

我已经尝试了一切，但我不知道该怎么做。 它给了我以下错误：

NotFoundError: 2 root error(s) found.
  (0) Not found: Resource localhost/total/N10tensorflow3VarE does not exist.
     [[{{node metrics/accuracy/AssignAddVariableOp}}]]
     [[metrics/precision/Mean/_87]]
  (1) Not found: Resource localhost/total/N10tensorflow3VarE does not exist.
     [[{{node metrics/accuracy/AssignAddVariableOp}}]]
0 successful operations.
0 derived errors ignored.

我将不胜感激任何帮助。 我是新来的。 我能做什么？ 谢谢！

Answer 1

似乎只有使用本机 keras 才会出现问题，但是当我尝试实现您的代码并在Tensorflow 2.x中对其进行如下修改时：

%tensorflow_version 2.x

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, Dense, Flatten
from tensorflow.keras.models import Model

METRICS = [
      keras.metrics.TruePositives(name='tp'),
      keras.metrics.FalsePositives(name='fp'),
      keras.metrics.TrueNegatives(name='tn'),
      keras.metrics.FalseNegatives(name='fn'), 
      keras.metrics.BinaryAccuracy(name='accuracy'),
      keras.metrics.Precision(name='precision'),
      keras.metrics.Recall(name='recall'),
      keras.metrics.AUC(name='auc'),
]

# function for creating a vgg block
def vgg_block(layer_in, n_filters, n_conv):
  # add convolutional layers
  for _ in range(n_conv):
    layer_in = Conv2D(n_filters, (3,3), padding='same', activation='relu')(layer_in)
  # add max pooling layer
  layer_in = MaxPooling2D((2,2), strides=(2,2))(layer_in)
  return layer_in

# define model input
visible = Input(shape=(256, 256, 3))
# add vgg module
layer = vgg_block(visible, 64, 2)

#####################################


flat = Flatten()(layer)
hidden1 = Dense(128, activation='relu')(flat)
output = Dense(1, activation='sigmoid')(hidden1)
model = Model(inputs=visible, outputs=output)
print(model.summary())

# # plot model architecture
# plot_model(model, show_shapes=True, to_file='vgg_block.png')

model.compile(loss='binary_crossentropy',
            optimizer='rmsprop',
            metrics=METRICS)

# New lines to obtain the best model in term of validation accuracy
from tensorflow.keras.callbacks import ModelCheckpoint
filepath="weights-improvement-{epoch:02d}-{val_accuracy:.2f}.h5"
checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
callbacks_list = [checkpoint]


## Synthetic Inputs
train_input = tf.random.normal((100, 256, 256, 3))
train_output = tf.random.normal((100, 1))

# Test Model.fit same as Model.fit_generator in TF 2.1.0
model.fit(train_input, train_output, epochs = 1)

问题没有出现，它工作正常。 你可以在 TF 2.x 中试试这个。 我希望这解决了你的问题。

Answer 2

问题出在 Keras 版本上。 尝试使用 Keras 2.3.0 及更高版本，它们引入了新的基于类的指标。

Keras 文档