TensorFlow 1.0 is promising more stable GPU build for Mac, let's test it.
Make sure you have CUDA installed
$ echo $DYLD_LIBRARY_PATH/usr/local/cuda/lib:
http://docs.nvidia.com/cuda/cuda-installation-guide-mac-os-x/#axzz4Ynrb42hi
$ echo $CUDA_HOME/Developer/NVIDIA/CUDA-8.0/
https://developer.nvidia.com/cudnn
$ sudo easy_install pip
Set variable with TensorFlow 1.0 binary
export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/mac/gpu/tensorflow_gpu-1.0.0-py3-none-any.whl
$ echo $TF_BINARY_URL https://storage.googleapis.com/tensorflow/mac/gpu/tensorflow_gpu-1.0.0-py3-none-any.whl
Use pip to install TensorFlow
$ sudo pip3 install --upgrade $TF_BINARY_URL Password:The directory '/Users/ukilucas/Library/Caches/pip/http' or its parent directory is not owned by the current user and the cache has been disabled. Please check the permissions and owner of that directory. If executing pip with sudo, you may want sudo's -H flag.The directory '/Users/ukilucas/Library/Caches/pip' or its parent directory is not owned by the current user and caching wheels has been disabled. check the permissions and owner of that directory. If executing pip with sudo, you may want sudo's -H flag.Collecting tensorflow-gpu==1.0.0 from https://storage.googleapis.com/tensorflow/mac/gpu/tensorflow_gpu-1.0.0-py3-none-any.whl Downloading https://storage.googleapis.com/tensorflow/mac/gpu/tensorflow_gpu-1.0.0-py3-none-any.whl (89.0MB) 100% |████████████████████████████████| 89.0MB 8.8kB/s Requirement already up-to-date: numpy>=1.11.0 in /Users/ukilucas/anaconda3/envs/tensorflow_gpu/lib/python3.5/site-packages (from tensorflow-gpu==1.0.0)Requirement already up-to-date: wheel>=0.26 in /Users/ukilucas/anaconda3/envs/tensorflow_gpu/lib/python3.5/site-packages (from tensorflow-gpu==1.0.0)Requirement already up-to-date: six>=1.10.0 in /Users/ukilucas/anaconda3/envs/tensorflow_gpu/lib/python3.5/site-packages (from tensorflow-gpu==1.0.0)Requirement already up-to-date: protobuf>=3.1.0 in /Users/ukilucas/anaconda3/envs/tensorflow_gpu/lib/python3.5/site-packages (from tensorflow-gpu==1.0.0)Collecting setuptools (from protobuf>=3.1.0->tensorflow-gpu==1.0.0) Downloading setuptools-34.2.0-py2.py3-none-any.whl (389kB) 100% |████████████████████████████████| 399kB 2.0MB/s Requirement already up-to-date: appdirs>=1.4.0 in /Users/ukilucas/anaconda3/envs/tensorflow_gpu/lib/python3.5/site-packages (from setuptools->protobuf>=3.1.0->tensorflow-gpu==1.0.0)Requirement already up-to-date: packaging>=16.8 in /Users/ukilucas/anaconda3/envs/tensorflow_gpu/lib/python3.5/site-packages (from setuptools->protobuf>=3.1.0->tensorflow-gpu==1.0.0)Requirement already up-to-date: pyparsing in /Users/ukilucas/anaconda3/envs/tensorflow_gpu/lib/python3.5/site-packages (from packaging>=16.8->setuptools->protobuf>=3.1.0->tensorflow-gpu==1.0.0)Installing collected packages: tensorflow-gpu, setuptools Found existing installation: tensorflow-gpu 0.12.1 Uninstalling tensorflow-gpu-0.12.1: Successfully uninstalled tensorflow-gpu-0.12.1 Found existing installation: setuptools 34.1.1 Uninstalling setuptools-34.1.1: Successfully uninstalled setuptools-34.1.1Successfully installed setuptools-34.2.0 tensorflow-gpu-1.0.0(tensorflow_gpu) uki@UkiPEsMcBookPro 192.168.1.24 19:44 dev $
Start jupyter notebook in the right (same) environment
(tensorflow_gpu) $ jupyter notebook
Try in jupyter notebook
import tensorflow as tf
import time
from tensorflow.python.client import device_lib
def get_available_CPU_GPU():
devices = device_lib.list_local_devices()
#return [x.name for x in devices if x.device_type == 'CPU']
return [x.name for x in devices ]
print(get_available_CPU_GPU())
['/cpu:0', '/gpu:0']
start = timeit.timeit()
print ("starting")with tf.device('/gpu:0'):
# [1.0, 2.0, 3.0, 4.0, 5.0, 6.0]
a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
c = tf.matmul(a, b)
# Creates a session with log_device_placement set to True.
sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))
# Runs the op.
print (sess.run(c))
end = timeit.timeit()
print ("elapsed", end - start)starting
[[ 22. 28.]
[ 49. 64.]]
elapsed -0.003607149003073573
It is a win so far, time will show if it is usable.
Reference:
https://www.tensorflow.org/install/install_mac
