/CogSci-WS2020-ANNs-in-Tensorflow

Primary LanguageJupyter Notebook

How to setup new conda environment for this repository:

conda create -n iannwtf python=3.7

-n or -name is the name option

How to activate environment to work with it:

conda activate iannwtf

command line starts with (iannwtf) then

conda install does work to install packages like TensorFlow, but we use pipinstall because typically pip is better kept up to date than conda for package updates.

Install pip inside of virtual conda environment

conda install pip

Get a list of all installed packages in our virtual environment:

conda list

Before you advance, please confirm that this list (should still be rather empty) contains pip now. We need to make sure of this, because when subsequently we use pip to install further packages, if we do not have pip in our virtual conda env, this will use your system’s pip to install python packages into your user, instead of the virtual env’s pip to install packages into your virtual env

seems to be not working pip install -upgrade pip therefore upgrade pip with:

pip install --upgrade pip

Installs packages and sub-dependencies

pip install tensorflow
pip install matplotlib

run test python script to verify tensorflow and matplotlib have been installed properly (assuming you are in the homework01 folder):

python .\test_tensorflow.py

Testing if GPU acceleration is working

import tensorflow as tf
from tensorflow.python.client import device_lib

tf.test.gpu_device_name()
device_lib.list_local_devices()

How to install conda for Linux Subsystem

wget https://repo.anaconda.com/archive/Anaconda3-20XX.XXXXXXXX-Linux-x86_64.sh
  • check the md5 checksum and run the downloaded file
md5sum Anaconda3-2019.03-Linux-x86_64.sh
bash Anaconda3-2019.03-Linux-x86_64.sh

remove shell script after installation

rm Anaconda3-2019.03-Linux-x86_64.sh

Now you should be able to start up your Anaconda environment:

source ~anaconda3/bin/activate

“~anaconda3/bin/activate” is default place that Anaconda will install itself but if you chose elsewhere simply point to that directory.

Once activated, initiate a full update:

conda update --all