Question

我是深度学习的新手。我试图在CPU上运行python的深度学习代码，效果很好，但相同的代码在使用gpu的tensorflow上不起作用。使用GPU的深度学习在语法上有什么区别吗？如果语法不同，那么任何入门的材料都会有所帮助。下面是在CPU上运行的用于二进制分类的简单代码，如果我想在GPU上运行它，我应该进行哪些必要的更改？

# Importing the Keras libraries and packages
from keras.models import Sequential
from keras.layers import Convolution2D
from keras.layers import MaxPooling2D
from keras.layers import Flatten
from keras.layers import Dense

# Initialising the CNN
classifier = Sequential()

# Step 1 - Convolution
classifier.add(Convolution2D(32, (3, 3), input_shape = (64, 64, 3),dilation_rate=(1,1), activation = 'relu', ))
classifier.add(Convolution2D(32, (3, 3),dilation_rate=(2,2), activation = 'relu', ))
classifier.add(Convolution2D(32, (3, 3),dilation_rate=(4,4), activation = 'relu', ))
#classifier.add(MaxPooling2D(pool_size = (2, 2)))

classifier.add(Convolution2D(64, (3, 3),dilation_rate=(1,1), activation = 'relu', ))
classifier.add(Convolution2D(64, (3, 3),dilation_rate=(2,2), activation = 'relu', ))
classifier.add(Convolution2D(64, (3, 3),dilation_rate=(4,4), activation = 'relu', ))



classifier.add(Convolution2D(128, (3, 3),dilation_rate=(1,1), activation = 'relu', ))
classifier.add(Convolution2D(128, (3, 3),dilation_rate=(2,2), activation = 'relu', ))
classifier.add(Convolution2D(128, (3, 3),dilation_rate=(4,4), activation = 'relu', ))


classifier.add(Convolution2D(256, (3, 3),dilation_rate=(1,1), activation = 'relu', ))
classifier.add(Convolution2D(256, (3, 3),dilation_rate=(2,2), activation = 'relu', ))
classifier.add(Convolution2D(256, (3, 3),dilation_rate=(4,4), activation = 'relu', ))

'''
classifier.add(Convolution2D(256, (3, 3),dilation_rate=(1,1), activation = 'relu', ))

#classifier.add(Convolution2D(512, (3, 3),dilation_rate=(2,2), activation = 'relu', ))
#classifier.add(Convolution2D(512, (3, 3),dilation_rate=(4,4), activation = 'relu', ))

classifier.add(Convolution2D(512, (3, 3),dilation_rate=(1,1), activation = 'relu', ))
#classifier.add(Convolution2D(1024, (3, 3),dilation_rate=(2,2), activation = 'relu', ))
#classifier.add(Convolution2D(1024, (3, 3),dilation_rate=(4,4), activation = 'relu', ))
'''

# Step 3 - Flattening
classifier.add(Flatten())

# Step 4 - Full connection
classifier.add(Dense(units = 256, activation = 'relu'))
classifier.add(Dense(units = 1, activation = 'sigmoid'))

# Compiling the CNN
classifier.compile(optimizer = 'adam', loss = 'binary_crossentropy', metrics = ['accuracy'])

# Part 2 - Fitting the CNN to the images

from keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(rescale = 1./255,
                                    featurewise_center=True,
                                    featurewise_std_normalization=True,
                                    rotation_range=20,
                                    width_shift_range=0.05,
                                    height_shift_range=0.05,
                                    shear_range = 0.05,
                                    zoom_range = 0.05,
                                    horizontal_flip = True)

test_datagen = ImageDataGenerator(rescale = 1./255)

training_set = train_datagen.flow_from_directory('Data_base/Processing_Data/Training',
                                                 target_size = (64, 64),
                                                 batch_size = 20,
                                                 class_mode = 'binary')

test_set = test_datagen.flow_from_directory('Data_base/Processing_Data/Test',
                                            target_size = (64, 64),
                                            batch_size = 6,
                                            class_mode = 'binary')

classifier.fit_generator(training_set,
                         samples_per_epoch =44 ,
                         nb_epoch = 20,
                         validation_data = test_set,
                         nb_val_samples =6 )
classifier.save_weights('first_try.h5')

Answer 1

您无需在代码中进行任何更改。

首先，如果要使用GPU，请确保已安装CUDA和cuDNN。您将需要的版本取决于您的GPU和TensorFlow版本。有一些教程。

第二个不要在同一环境中安装tensorflow和tensorflow-gpu。至少对我来说，这引起了一些奇怪的错误。（我不知道这是否已经解决。）

使用和不使用GPU编程的语法差异？

1 个答案: