rlan

1. Create virtual environment

virtualenv -p /usr/local/bin/python3 my_project
source my_project/bin/activate

2. Install tensorflow
3. Install the following in the virtual environment:

rlan

Global

import logger
def setupLogging():
    """
    setup logging
    """
    logger = logging.getLogger('app')
 logger.setLevel(logging.DEBUG)

rlan

MNIST	Plain	AVX+FMA+SSE	Intel MKL-DNN
Train (s)	83.375	64.026	77.266
Evaluation (s)	3.50	3.47	6.107

• Tensorflow 1.9.0
• Macbook Pro 13", 2016, 2.4 GHz Intel Core i7, 16 GB RAM, macOS Sierra

import tensorflow as tf

rlan

nvidia-smi -q | grep Product

Product Name                    : GeForce GTX 1080 Ti
Product Brand                   : GeForce
Product Name                    : GeForce GTX 1080 Ti
Product Brand                   : GeForce
Product Name                    : GeForce GTX 1080 Ti
Product Brand                   : GeForce
Product Name : GeForce GTX 1080 Ti

rlan

See: https://docs.nvidia.com/deploy/cuda-compatibility/index.html#binary-compatibility__table-toolkit-driver

Original: https://stackoverflow.com/questions/30820513/what-is-version-of-cuda-for-nvidia-304-125/30820690#annotations:JZkItJrNEeivgxOuB7-5Vg

rlan

nvcc --version

rlan

You can do it faster without any imports just by using magics:

%env CUDA_DEVICE_ORDER=PCI_BUS_ID
%env CUDA_VISIBLE_DEVICES=0

Notice that all env variable are strings, so no need to use ". You can verify that env-variable is set up by running: %env <name_of_var>. Or check all of them with %env.

rlan

Now you can tell Git not to guess, but rather to insist that you set user.name and user.email explicitly before it will let you commit:

git config --global user.useconfigonly true

https://blog.github.com/2016-03-28-git-2-8-has-been-released/#annotations:eoCI6HhFEeiNcDNxG8M_Fg

rlan

Chooses gpu:0 or the first GPU in a GPU cluster.

CUDA_VISIBLE_DEVICES=0 python deeplab/model_test.py

More info about this environment variable: https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#env-vars

rlan

Some functions return the self object. In this case, the object is an instance at an address. We want to ignore this output:

I found it easier to simply assign the unneeded return values to a variable:

>>> _ = do_something()
>>> check_something()
True

https://stackoverflow.com/a/22537703/1704566