fermaat / afi_deep_learning_intro

Software instalation and previous steps

Linux machines (ubuntu)

You need to install pip3 on your system

sudo apt install python3-pip

Instal virtualenv

pip3 install virtualenv

Create environment

virtualenv path_to_virtual_envs_location/environment_name

activate virtuelenv (if needed)

source path_to_virtual_envs_location/environment_name/bin/activate

install dependences pip

pip3 install -r requirements.txt

install dependences

pip3 install -r requirements.txt In case you have cuda installed and configured, just replace tensorflow package with tensorflow-gpu:

pip3 install -r requirements_gpu.txt

Jupyter kernel generation

After having sourced the environment:

ipython kernel install --user --name=environment_name

Windows machines

First of all, it is required to have a python v 3.6.8 with 64 bits. It can be obtained from here. Probably another pthon versions would work, but 3.8+ has proven to have an issue with pip and tensorflow.

This version includes pip, so it is not necessary to install it.

Scripts enabling

By default, Powershell does not enable you to execute scripts. So, It is necessary t enable them.

This command will do, according to this thread

Set-ExecutionPolicy -ExecutionPolicy Unrestricted-Scope CurrentUser

Environment creation and activation

It is recommended to run everything on an environment

pip install virtualenv

Create virtualenv pip

virtualenv path_to_virtual_envs_location/environment_name

Create virtualenv conda

conda create -n environment_name

And in order to activate the virtualenv: conda activate environment_name

Activate virtualenv pip

This is why we enabled the script execution for the current user. If we are using cmd, just run this command

./path_to_virtual_envs_location/environment_name/Scripts/activate.bat If we are using powershell,

./path_to_virtual_envs_location/environment_name/Scripts/activate.ps1

Install requirements

Now we are sourcing our created environment. Just install the requirements:

pip install -r requirements_windows.txt

Notebook installation

Although it should be contained on the requirements, it is a complete different requirement than on ubuntu. So, it is needed to run this line: pip install notebook

Torch

If you want also to install torch, you need to run this line (only for cpu)

pip install torch==1.3.1+cpu torchvision==0.4.2+cpu -f https://download.pytorch.org/whl/torch_stable.html

Jupyter kernel generation

After having sourced the environment:

ipython kernel install --user --name=environment_name