头哥装Tensorflow-gpu

Tensorflow-gpu  版本1.4

1）  pip install -i https://pypi.tuna.tsinghua.edu.cn/simple tensorflow-gpu==1.4.0

2） CUDA® Toolkit 8.0， 需要注意最新版9.1不支持tensorflow 1.4版本；

3）   去 nv开发者注册    https://developer.nvidia.com/cudnn （艰难！  ）装 cuDNN v6.0 Library for Windows 10

4） 覆盖  C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\     

执行下面的 python  ，出现 gpu 字样 即可

import numpy as np

np.random.seed(1337) 

from keras.datasets import mnist

from keras.utils import np_utils

from keras.models import Sequential

from keras.layers import Dense, Activation

from keras.optimizers import RMSprop

# 下载数据集

(X_train, y_train), (X_test, y_test) = mnist.load_data()

# 数据预处处理

X_train = X_train.reshape(X_train.shape[0], -1) / 255.

X_test = X_test.reshape(X_test.shape[0], -1) / 255. 

y_train = np_utils.to_categorical(y_train, num_classes=10)

y_test = np_utils.to_categorical(y_test, num_classes=10)

# 不使用model.add()，用以下方式也可以构建网络

model = Sequential([

    Dense(400, input_dim=784),

    Activation('relu'),

    Dense(10),

    Activation('softmax'),

])

# 定义优化器

rmsprop = RMSprop(lr=0.001, rho=0.9, epsilon=1e-08, decay=0.0)

model.compile(optimizer=rmsprop,

              loss='categorical_crossentropy',

              metrics=['accuracy']) # metrics赋值为'accuracy'，会在训练过程中输出正确率

# 这次我们用fit()来训练网路

print('Training ------------')

model.fit(X_train, y_train, epochs=4, batch_size=32)

print('\nTesting ------------')

# 评价训练出的网络

loss, accuracy = model.evaluate(X_test, y_test)

print('test loss: ', loss)

print('test accuracy: ', accuracy)