Keras Env¶
Keras comes packaged with TensorFlow 2.0 as tensorflow.keras.
To start using Keras, simply install TensorFlow 2.0.
- Keras/Tensorflow 兼容的设备
- Python 3.5-3.8
- Ubuntu 16.04 or later
- Windows 7 or later
- macOS 10.12.6(Sierra) or later
- 在安装 Keras 之前, 需要安装一个后端引擎: TensorFlow, Theano, CNTK
- 安装可选依赖项
- cuDNN(如果您计划在GPU上运行Keras, 建议使用)
- HDF5和h5py(如果您计划将Keras模型保存到磁盘, 则需要)
- graphviz和pydot(由可视化实用程序用于绘制模型图)
- 安装 Keras
- 配置 Keras 后端
1. 安装后端引擎¶
Tensorflow 2.0:
# Tensorflow GPU version
pip install --upgrade tensorflow-gpu
# Tensorflow CPU version
pip install --upgrade tensorflow
Theano:
$ todo
CNTK:
$ todo
2. 安装可选依赖项¶
cUNN:
HDF5,h5py:
$ conda install h5py
or
$ pip install h5py
graphviz:
$ pip install
pydot:
$ pip install
3. 安装 Keras¶
从 PYPI 安装 Keras:
# keras nonvirtualenv version
$ sudo pip install keras
# keras virtualenv version
$ pip install keras
# tensorflow version
$ pip install keras -U --pre
从 GitHub 源安装 Keras:
$ cd /usr/local
$ git clone https://github.com/keras-team/keras.git
$ cd keras
$ sudo python setup.py install