intro-ml-python

Playground project for learning purposes

setup

first time

conda create --name tf pip python=3.6 #create virtual environment
source activate tf  # enter created environment
pip install -r requirements.txt # inside environment install tensorflow and all dependencies

next time

source activate tf  # enter created environment
# or 
conda activate tf

occasionally

pip install --upgrade tensorflow
pip freeze > requirements.txt 
pip install -r requirements.txt
pip install -r requirements.txt --upgrade
conda update conda
pip install --upgrade pip
pip install --upgrade jupyter
pip install --upgrade notebook
pip install --upgrade tensorboard

jupyter notebook #start jupyter

conda update -n base -c defaults conda
 

tensor board

tensorboard --logdir func_approx/.tensorboard-func_approx.py

Running from commandline

• Make sure tf environment is active
• Use run.sh <path_to_python_script>, for example

./run.sh func_approx/func_approx.py

• or add this to the root of the python script:

import os,sys,inspect
currdir = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe())))
sys.path.append(os.path.dirname(currdir))

modules and auto reloading

tl/df

%load_ext autoreload
%autoreload 2
# %aimport

https://ipython.readthedocs.io/en/stable/config/extensions/autoreload.html

Setup IDE

Find the environment path:

conda info --envs | grep tf

Intellij Ultimate Edition

Open project settings and add python SDK pointing to proper virtual environment found above:

And mark folder as sources.

PyCharm Community Edition

Activation functions