尝试训练 TensorFlow 模型时如何修复 InvalidArgument 错误？

Question

我正在为猫和狗图像分类项目训练机器学习模型，但在尝试训练模型时似乎遇到了错误，我认为这可能是由于我传递的输入数据类型或模型结构中的缺陷。

创建目录、预处理和数据集变量：

# Get project files
!wget https://cdn.freecodecamp.org/project-data/cats-and-dogs/cats_and_dogs.zip
!unzip cats_and_dogs.zip
PATH = 'cats_and_dogs'
train_dir = os.path.join(PATH, 'train')
validation_dir = os.path.join(PATH, 'validation')
test_dir = os.path.join(PATH, 'test')
# Get number of files in each directory. The train and validation directories
# each have the subdirecories "dogs" and "cats".
total_train = sum([len(files) for r, d, files in os.walk(train_dir)])
total_val = sum([len(files) for r, d, files in os.walk(validation_dir)])
total_test = len(os.listdir(test_dir))
# Variables for pre-processing and training.
batch_size = 128
epochs = 15
IMG_HEIGHT = 150
IMG_WIDTH = 150

train_image_generator = tf.keras.preprocessing.image.ImageDataGenerator(rescale=0.00392156862)
validation_image_generator = tf.keras.preprocessing.image.ImageDataGenerator(rescale=0.00392156862)
test_image_generator = tf.keras.preprocessing.image.ImageDataGenerator(rescale=0.00392156862)

train_data_gen = train_image_generator.flow_from_directory(directory=train_dir, target_size=(IMG_HEIGHT,IMG_WIDTH),class_mode='binary',batch_size=batch_size,shuffle=True,color_mode='rgb')
val_data_gen = validation_image_generator.flow_from_directory(directory=validation_dir, target_size=(IMG_HEIGHT,IMG_WIDTH),class_mode='binary',batch_size=batch_size,shuffle=True,color_mode='rgb')
test_data_gen = test_image_generator.flow_from_directory(directory=PATH,target_size=(IMG_HEIGHT,IMG_WIDTH),classes=['test'],class_mode='input',batch_size=batch_size,shuffle=False,color_mode='rgb')

构建和训练模型：

model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Conv2D(32,(3,3),activation='relu',input_shape=(32,32,3)))
model.add(tf.keras.layers.MaxPooling2D((2,2)))
model.add(tf.keras.layers.Conv2D(64,(3,3),activation='relu'))
model.add(tf.keras.layers.MaxPooling2D((2,2)))
model.add(tf.keras.layers.Conv2D(64,(3,3),activation='relu'))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(32))
model.add(tf.keras.layers.Dense(16))
model.add(tf.keras.layers.Dense(1, activation='sigmoid'))
model.summary()
model.compile(optimizer="Adam",loss='binary_crossentropy',metrics=['accuracy'])
#training
history = model.fit(train_data_gen,steps_per_epoch=round(total_train)//batch_size,epochs=epochs,batch_size=batch_size)

错误截图： InvalidArgumentError

关于什么可能导致问题的任何想法？

Answer 1

在您的生成器中，您指定 IMG_HEIGHT = 150 和 IMG_WIDTH = 150。但是在您的模型中，您有代码

model.add(tf.keras.layers.Conv2D(32,(3,3),activation='relu',input_shape=(32,32,3)))

将输入形状更改为 (150, 150,3)。作为旁注，你编码

train_image_generator = tf.keras.preprocessing.image.ImageDataGenerator(rescale=0.00392156862)
validation_image_generator = tf.keras.preprocessing.image.ImageDataGenerator(rescale=0.00392156862)
test_image_generator = tf.keras.preprocessing.image.ImageDataGenerator(rescale=0.00392156862)

因为它们是相同的，所以只需使用

gen=tf.keras.preprocessing.image.ImageDataGenerator(rescale=0.00392156862)

然后

train_data_gen = gen.flow_from_directory(directory=train_dir, target_size=(IMG_HEIGHT,IMG_WIDTH),class_mode='binary',batch_size=batch_size,shuffle=True,color_mode='rgb')
val_data_gen = gen.flow_from_directory(directory=validation_dir, target_size=(IMG_HEIGHT,IMG_WIDTH),class_mode='binary',batch_size=batch_size,shuffle=True,color_mode='rgb')
test_data_gen = gen.flow_from_directory(directory=PATH,target_size=(IMG_HEIGHT,IMG_WIDTH),classes=['test'],class_mode='input',batch_size=batch_size,shuffle=False,color_mode='rgb')

同样在 model.fit 中，因为您使用的是生成器，所以您不应该包含参数 batch_size，因为它已经在生成器中指定。我也不包括参数steps_per_epoch。如果您不包含它，那么 model.fit 将在内部正确计算它。