TensorFlow Build

Build of TensorFlow Master, 1.8 and 1.7.1 on Docker with Ubuntu 16.04, GCC 5.4, Python 3.6, CUDA 9.1 and cuDNN 7.1.

(old build documentation for TensorFlow 0.10 in Portuguese)

http://cirocavani.github.io/post/compilacao-do-tensorflow-0.10-para-linux-com-gpu/

...

Setup

Manual download required:

• _deps/cudnn-9.1-linux-x64-v7.1.tgz: cuDNN 7.1 for CUDA 9.1 [ Download ]

./build_image.sh cpu
./build_image.sh gpu

Builds

• ./build_tensorflow.sh master gpu: builds package master/tensorflow_gpu-1.9.0a0-cp36-cp36m-linux_x86_64.whl
• ./build_tensorflow.sh master cpu: builds package master/tensorflow-1.9.0a0-cp36-cp36m-linux_x86_64.whl
• ./build_tensorflow.sh 1.8 gpu: builds package 1.8/tensorflow_gpu-1.8.0-cp36-cp36m-linux_x86_64.whl
• ./build_tensorflow.sh 1.8 cpu: builds package 1.8/tensorflow-1.8.0-cp36-cp36m-linux_x86_64.whl
• ./build_tensorflow.sh 1.7 gpu: builds package 1.7/tensorflow_gpu-1.7.1-cp36-cp36m-linux_x86_64.whl
• ./build_tensorflow.sh 1.7 cpu: builds package 1.7/tensorflow-1.7.1-cp36-cp36m-linux_x86_64.whl

Install

# Python install
curl -k -L -O https://repo.continuum.io/miniconda/Miniconda3-4.5.1-Linux-x86_64.sh
bash Miniconda3-4.5.1-Linux-x86_64.sh -b -f -p tensorflow-gpu
tensorflow-gpu/bin/conda update -y -q --all

# TensorFlow install
tensorflow-gpu/bin/pip install 1.8/tensorflow_gpu-1.8.0-cp36-cp36m-linux_x86_64.whl

tensorflow-gpu/bin/python -c \
'import tensorflow as tf; print(tf.__version__); print(tf.Session().run(tf.constant("Hello, World!")))'
#> 1.8.0
#> b'Hello, World!'