Question

我正在使用tf.keras训练具有以下摘要的模型：

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (None, 28, 28, 20)        520       
_________________________________________________________________
flatten (Flatten)            (None, 15680)             0         
_________________________________________________________________
dense (Dense)                (None, 5)                 78405     
=================================================================
Total params: 78,925
Trainable params: 78,925
Non-trainable params: 0
_________________________________________________________________

我正在使用

调用fit方法

model.fit(X_train, y_train, batch_size=32, steps_per_epoch=125, epochs=5, use_multiprocessing=True)

其中X_train是形状为[900000,32,32,1]的张量流变量。

我遇到以下错误：

Resource exhausted: OOM when allocating tensor with shape[900000,28,28,20] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
     [[{{node conv2d_1/Conv2D}}]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

     [[metrics_2/acc/Identity/_53]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

  (1) Resource exhausted: OOM when allocating tensor with shape[900000,28,28,20] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
     [[{{node conv2d_1/Conv2D}}]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

0 successful operations.
0 derived errors ignored.

当批次大小为32时，我无法理解为什么它会分配形状为[900000,28,28,20]的张量。我期望的是[32,28,28,20]。

完整代码：

IMAGE_SIZE = 32
BATCH_SIZE = 32

conv1_layer = keras.layers.Conv2D(input_shape=(IMAGE_SIZE, IMAGE_SIZE, 1), filters=20, kernel_size=[5, 5], activation='relu')
f = keras.layers.Flatten()
output_layer = keras.layers.Dense(units=5, activation='softmax')

model = keras.models.Sequential(layers=[conv1_layer, f, output_layer])
model.compile(optimizer='sgd', loss='categorical_crossentropy', metrics=['accuracy'])
model.fit(X_train, y_train, batch_size=BATCH_SIZE, steps_per_epoch=int(X_train.shape[0])//BATCH_SIZE, epochs=5, use_multiprocessing=True)

Answer 1

该问题最可能出现在use_multiprocessing=True中，该问题期望输入作为生成器对象（请参见docs）-因此，如果馈送X_train数组不会立即产生错误，则可能通过在轴0上进行迭代并一次馈送所有900,000个样本，将其视为发生器，从而产生误差。尝试使用use_multiprocessing=False，或使用generator输入数据。

此外，steps_per_epoch可能是一个附加来源-从fit中省略它们-然后，在输入tf.Variable之前，先评估，例如{{ 1}}不会自动处理。

Answer 2

我建议您编写自己的训练数据生成器或keras.utils.Sequence。直接输入整个训练数据不是一个好主意，特别是当您的数据集非常大时。另外， use_multiprocessing自变量用于generator或keras.utils。仅按顺序输入，如.fit() doc

中所述

以下代码改编自A detailed example of how to use data generators with Keras。

import numpy as np
import keras

class DataGenerator(keras.utils.Sequence):
    'Generates data for Keras'
    def __init__(self, list_IDs, labels, batch_size=32, dim=(32,32), n_channels=1, n_classes=10, shuffle=True):
        'Initialization'
        self.dim = dim
        self.batch_size = batch_size
        self.labels = labels
        self.list_IDs = list_IDs
        self.n_channels = n_channels
        self.n_classes = n_classes
        self.shuffle = shuffle
        self.on_epoch_end()

    def __len__(self):
        'Denotes the number of batches per epoch'
        return int(np.floor(len(self.list_IDs) / self.batch_size))

    def __getitem__(self, index):
        'Generate one batch of data'
        # Generate indexes of the batch
        indexes = self.indexes[index*self.batch_size:(index+1)*self.batch_size]

        # Find list of IDs
        list_IDs_temp = [self.list_IDs[k] for k in indexes]

        # Generate data
        X, y = self.__data_generation(list_IDs_temp)

        return X, y

    def on_epoch_end(self):
        'Updates indexes after each epoch'
        self.indexes = np.arange(len(self.list_IDs))
        if self.shuffle == True:
            np.random.shuffle(self.indexes)

    def __data_generation(self, list_IDs_temp):
        'Generates data containing batch_size samples' # X : (n_samples, *dim, n_channels)
        # Initialization
        X = np.empty((self.batch_size, *self.dim, self.n_channels))
        y = np.empty((self.batch_size), dtype=int)

        # Generate data
        for i, ID in enumerate(list_IDs_temp):
            # Store sample
            X[i,] = np.load('data/' + ID + '.npy')

            # Store class
            y[i] = self.labels[ID]

        return X, keras.utils.to_categorical(y, num_classes=self.n_classes)

params = {'dim': (32,32),
          'batch_size': 32,
          'n_classes': 5,
          'n_channels': 1,
          'shuffle': True}

train_data_path = # (list of string) ['0302','0325','3103',...] 'data/0302.npy' is the path where stores the '0302' image. 
train_labels = # (list of int) [1,0,4,...] corresponding labels of 900,000 images.

# Generators
train_gen = DataGenerator(train_data_path, train_labels, **params)

model.fit_generator(generator=train_gen,
                    epochs=5,
                    use_multiprocessing=True,
                    workers=2)

ResourceExhaustedError：在Keras中分配张量时，OOM

2 个答案: