Main repo for system design and playing around with different architectures
We recommend you install Anaconda to have a fixed environment for your OpenAI gym, tensorflow and other libarary for your dev environment.
You can download binary distribution of anaconda and install it for your preferred OS at: https://www.anaconda.com/distribution/
If you don't want Anaconda to activate the "base" environment by default, run this command:
conda config --set auto_activate_base false
conda create -n gym python=3 pip
This creates an environment called "gym", now you can switch and install gym package
conda activate gym
OpenAI Gym contains some RL environments that you can experiment with.
pip install gym
First you need to install the rl_salk gym env
cd libs
pip install -e .
You can now import and create the installed env
python
import gym
import rl_salk
env = gym.make('grid-world-v0')
The docs: http://gym.openai.com/docs/
Blog post: https://medium.com/@apoddar573/making-your-own-custom-environment-in-gym-c3b65ff8cdaa
Jupyter notebook offers a nice interface for putting code, numerical results, and graphics in the same document. Generally you interact with Jupyter notebooks by running a local server on your computer (jupyter notebook) and pointing your browser to the appropriate page. After you have created a notebook locally, if you add it to a Git repository on GitHub, you can view the notebook on GitHub without running a server (e.g. the Cliff Walk notebook).
conda activate <your-env-name>
pip install jupyter ipython
cd notebooks/
jupyter notebook
If you create a new environment that is different from the default one, you need to install ipythonkernel in the new environments. Here is how I did:
In the terminal (replacing <your-env-name> with gym, for instance):
conda activate <your-env-name>
conda install ipykernel
python -m ipykernel install --user --name <your-env-name> --display-name "Python (<your-env-name>)"
jupyter notebook
In the notebook, click: Kernel -> Change Kernel -> Python ()
This procedure is from: How to install IPython Kernels for different environments