There are many ways to install TensorFlow. This article will provide a detailed guide on how to install TensorFlow using pip.
Available Installation Packages
-
tensorflow — Current version for CPU only (recommended for beginners)
-
tensorflow-gpu — Current version with GPU support (Ubuntu and Windows)
-
tf-nightly — Nightly build for CPU only (unstable)
-
tf-nightly-gpu — Nightly build with GPU support (unstable, Ubuntu and Windows)
System Requirements
-
Ubuntu 16.04 or later (64-bit)
-
macOS 10.12.6 (Sierra) or later (64-bit) (no GPU support)
-
Windows 7 or later (64-bit) (Python 3 only)
-
Raspbian 9.0 or later
Hardware Requirements
-
Starting from TensorFlow 1.6, binaries use AVX instructions, which may not run on older CPUs
-
Refer to the GPU support guide (https://tensorflow.google.cn/install/gpu?hl=zh-CN) to set up a CUDA® supported GPU card on Ubuntu or Windows
Installing Python Development Environment on Your System
Python 3
Check if your Python environment is configured:
Python 3.4, 3.5, or 3.6 is required
$ python3 –version$ pip3 –version$ virtualenv –version
If these packages are installed, proceed to the next step.
Otherwise, install Python, pip package manager, and Virtualenv:
UBUNTU
$ sudo apt update$ sudo apt install python3-dev python3-pip$ sudo pip3 install -U virtualenv # system-wide install
MAC OS
Install using Homebrew package manager:
$ /usr/bin/ruby -e “$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)”$ export PATH=”/usr/local/bin:/usr/local/sbin:$PATH”$ brew update$ brew install python # Python 3$ sudo pip3 install -U virtualenv # system-wide install
WINDOWS
Install 2015 Redistributable Update 3, which comes with Visual Studio 2015 and can be installed separately:
-
Go to Visual Studio downloads
-
Select Redistributables and Build Tools
-
Download and install Microsoft Visual C++ 2015 Redistributable Update 3
Install the 64-bit Python 3 distribution for Windows (select pip as an optional feature)
C:\> pip3 install -U pip virtualenv
RASPBERRY PI
$ sudo apt update$ sudo apt install python3-dev python3-pip$ sudo apt install libatlas-base-dev # required for numpy$ sudo pip3 install -U virtualenv # system-wide install
OTHER
$ curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py$ python get-pip.py$ sudo pip3 install -U virtualenv # system-wide install
Python 2.7
Check if your Python environment is configured:
$ python –version$ pip –version$ virtualenv –version
If these packages are installed, proceed to the next step.
Otherwise, install Python, pip package manager, and Virtualenv:
UBUNTU
$ sudo apt update$ sudo apt install python-dev python-pip$ sudo pip install -U virtualenv # system-wide install
MAC OS
Install using Homebrew package manager:
$ /usr/bin/ruby -e “$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)”$ export PATH=”/usr/local/bin:/usr/local/sbin:$PATH”$ brew update$ brew install python@2 # Python 2$ sudo pip install -U virtualenv # system-wide install
RASPBERRY PI
$ sudo apt update$ sudo apt install python-dev python-pip$ sudo apt install libatlas-base-dev # required for numpy$ sudo pip install -U virtualenv # system-wide install
OTHER
$ curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py$ python get-pip.py$ sudo pip install -U virtualenv # system-wide install
Creating a Virtual Environment (Recommended)
Python virtual environments are used to isolate package installations from the system.
UBUNTU / MAC OS
Create a new virtual environment by selecting the Python interpreter and creating a ./venv directory:
$ virtualenv –system-site-packages -p python2.7 ./venv
Activate the virtual environment using shell-specific commands:
$ source ./venv/bin/activate # sh, bash, ksh, or zsh
When virtualenv is active, the shell prompt is prefixed with (venv).
To install packages in the virtual environment without affecting the host system settings, first upgrade pip:
(venv)$ pip install –upgrade pip(venv)$ pip list # show packages installed within the virtual environment
Then deactivate virtualenv:
(venv)$ deactivate # don’t exit until you’re done using TensorFlow
WINDOWS
Create a new virtual environment by selecting the Python interpreter and creating a ./venv directory.
Activate the virtual environment:
(venv)C:\> .envin\activate
Install packages in the virtual environment without affecting the host system settings. First upgrade pip:
(venv)C:\> pip install –upgrade pip(venv)C:\> pip list # show packages installed within the virtual environment
Then deactivate virtualenv:
(venv)C:\> deactivate # don’t exit until you’re done using TensorFlow
CONDA
We recommend using the pip package provided by TensorFlow, but you can also use the community-supported Anaconda package.
Create a new virtual environment by selecting the Python interpreter and creating a ./venv directory:
$ conda create -nvenv pip python=2.7
Activate the virtual environment:
& source activate venv
Within the virtual environment, install the TensorFlow pip package using its full URL:
(venv)$ pip install –ignore-installed –upgrade packageURL
Then deactivate virtualenv:
(venv)$ source deactivate
Installing the TensorFlow pip Package
Select one of the following TensorFlow packages from PyPI for installation:
-
tensorflow — Current version for CPU only (recommended for beginners)
-
tensorflow-gpu — Current version with GPU support (Ubuntu and Windows)
-
tf-nightly — Nightly build for CPU only (unstable)
-
tf-nightly-gpu — Nightly build with GPU support (unstable, Ubuntu and Windows)
Package dependencies are installed automatically. They are all listed under REQUIRED_PACKAGES in the setup.py file.
VIRTUALENV INSTALL
(venv)$ pip install –upgrade tensorflow
Verify the installation:
(venv)$ python -c “import tensorflow as tf; print(tf.__version__)”
SYSTEM INSTALL
$ pip install –user –upgrade tensorflow # install in $HOME
Verify the installation:
$ python -c “import tensorflow as tf; print(tf.__version__)”
Success: TensorFlow is now installed. Read the tutorial to get started. (https://tensorflow.google.cn/tutorials/?hl=zh-CN)
Package Locations
Some installation mechanisms require the URL of the TensorFlow Python package. The value you specify depends on your Python version.
|
Version |
URL |
|
Linux |
|
|
Python 2.7 CPU-only |
https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.11.0-cp27-none-linux_x86_64.whl |
|
Python 2.7 GPU support |
https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.11.0-cp27-none-linux_x86_64.whl |
|
Python 3.4 CPU-only |
https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.11.0-cp34-cp34m-linux_x86_64.whl |
|
Python 3.4 GPU support |
https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.11.0-cp34-cp34m-linux_x86_64.whl |
|
Python 3.5 CPU-only |
https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.11.0-cp35-cp35m-linux_x86_64.whl |
|
Python 3.5 GPU support |
https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.11.0-cp35-cp35m-linux_x86_64.whl |
|
Python 3.6 CPU-only |
https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.11.0-cp36-cp36m-linux_x86_64.whl |
|
Python 3.6 GPU support |
https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.11.0-cp36-cp36m-linux_x86_64.whl |
|
macOS (CPU-only) |
|
|
Python 2.7 |
https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.11.0-py2-none-any.whl |
|
Python 3.4, 3.5, 3.6 |
https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.11.0-py3-none-any.whl |
|
Windows |
|
|
Python 3.5 CPU-only |
https://storage.googleapis.com/tensorflow/windows/cpu/tensorflow-1.11.0-cp35-cp35m-win_amd64.whl |
|
Python 3.5 GPU support |
https://storage.googleapis.com/tensorflow/windows/gpu/tensorflow_gpu-1.11.0-cp35-cp35m-win_amd64.whl |
|
Python 3.6 CPU-only |
https://storage.googleapis.com/tensorflow/windows/cpu/tensorflow-1.11.0-cp36-cp36m-win_amd64.whl |
|
Python 3.6 GPU support |
https://storage.googleapis.com/tensorflow/windows/gpu/tensorflow_gpu-1.11.0-cp36-cp36m-win_amd64.whl |
