docker 下keras & tensorflow的一些问题汇总（待续）

1. 下载keras镜像之后，运行keras相关的代码一异常抛出

  File "<stdin>", line 1, in <module>

  File "/root/miniconda2/lib/python2.7/site-packages/keras/__init__.py", line 3, in <module>

    from . import activations

  File "/root/miniconda2/lib/python2.7/site-packages/keras/activations.py", line 3, in <module>

    from . import backend as K

  File "/root/miniconda2/lib/python2.7/site-packages/keras/backend/__init__.py", line 73, in <module>

    from .tensorflow_backend import *

  File "/root/miniconda2/lib/python2.7/site-packages/keras/backend/tensorflow_backend.py", line 6, in <module>

    from tensorflow.python.ops import ctc_ops as ctc

ImportError: cannot import name ctc_ops

解决办法：

pip install tensorflow --upgrade

其它：~/.keras/keras.json下面有配置backend的

Keras后端

什么是“后端”

Keras是一个模型级的库，提供了快速构建深度学习网络的模块。Keras并不处理如张量乘法、卷积等底层操作。这些操作依赖于某种特定的、优化良好的张量操作库。Keras依赖于处理张量的库就称为“后端引擎”。Keras提供了两种后端引擎Theano/Tensorflow，并将其函数统一封装，使得用户可以以同一个接口调用不同后端引擎的函数

Theano是一个开源的符号主义张量操作框架，由蒙特利尔大学LISA/MILA实验室开发

TensorFlow是一个符号主义的张量操作框架，由Google开发

2. docker 镜像有东西没有vim 等操作需要安装

apt-get install vim 的时候，找不到可用的vim

则需要apt-get update 更新source.list

3.  docker一些常用 的命令

docker run -i -t -v /data/software/:/mnt/software/ image_name /bin/bash

docker search image_name

docker pull image_name

docker start
container_id

docker attach
container_id

docker rm
container_id

docker exec
container_id

docker commit container_id image_name

from keras.models import Sequential

from keras.layers import Dense, Dropout, Activation

from keras.optimizers import SGD

model = Sequential()

# Dense(64) is a fully-connected layer with 64 hidden units.

# in the first layer, you must specify the expected input data shape:

# here, 20-dimensional vectors.

model.add(Dense(64, activation='relu', input_dim=20))

model.add(Dropout(0.5))

model.add(Dense(64, activation='relu'))

model.add(Dropout(0.5))

model.add(Dense(10, activation='softmax'))

sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)

model.compile(loss='categorical_crossentropy',

              optimizer=sgd,

              metrics=['accuracy'])

x_train = np.random.random((1000, 20))

y_train = np.random.random((1000, 10))

model.fit(x_train, y_train,

          epochs=20,

          batch_size=128)

score = model.evaluate(x_test, y_test, batch_size=128)