Some of the training tutorials you will find in this section will be using TensorFlow.js to train models and some others will be using original version of Tensorflow written in Python.
When working with Python, we recommend you to install a package dependency and environment management tool, like Anaconda. This allows to manage your different Python versions and dependencies.
To install Conda please follow this instructions: (Thanks to shekit!)
1) Install miniconda
- Go to https://conda.io/miniconda.html
- Choose Python 3.6 (bash installer) and download
- Open terminal and type:
(For the path you can drag the bash file you download into your terminal window from where you installed it.)
- Review the license and approve the license terms - type in
yesand press enter
Enteragain to confirm the location of install
yeswhen it asks you if the install location should be prepended to PATH
- Restart Terminal for changes to take effect
- If it prints out some stuff then it has installed correctly
2) Create an environment
conda create -n ml5 python=3.5.2
You can name it something other than 'ml5' if you prefer. Type:
y (and press Enter). This will create a conda environment with the name 'ml5' and python version 3.5.2
3) Turn off conda by default
The above instructions will set conda to be your "default" python on your machine (rather than the usual python 2 that comes pre-installed on a mac.) If you would prefer to turn this off, you have to edit your
bash_profile (a configuration file for terminal.) Use these steps.
Edit bash profile with
You should see:
# added by Miniconda3 4.3.11 installer export PATH="/Users/yourname/miniconda3/bin:$PATH"
Change this to:
Restart terminal. Now terminal will not use your conda python installation unless you enter
start_conda. You could also consider using something like VirtualEnv instead.
4) Activate environment
source activate ml5
You should see (ml5) prepended before your terminal prompt
5) Install python packages
Create a file called
requirements.txt and paste the following into it.
numpy==1.10.4 scipy==0.17.0 tensorflow==1.0.0
pip install -r requirements.txt
Make sure you
ml5 environment is activated (you should see (ml5) prepended before your terminal prompt).
All set! 🌈