codinalot / policy-gradient-example

This is just an example script

PyTorch Policy Gradient Cart-Pole Solver 🐍🛒

Welcome to the PyTorch Policy Gradient Cart-Pole Solver repository! This repository contains two scripts that offer exciting functionalities.

policygrad.py 📃

• Trains a multi-layer perceptron as a policy model for the cartpole environment.
• Allows you to watch the model play a game in the cartpole OpenAI Gym environment by setting the watch_play flag.
• Plots the rewards over time when the plot_rewards flag is set.
• Invokes the viz_conversion script to store neurons, synapses, and weights for each episode in JSON format.

viz_conversion.py 📈

• Converts the recorded neurons, synapses, edges, and weights for each episode into a JSON document.
• Converts the rewards for each episode into a separate JSON document.
• Produces the model.json and rewards.json output files.

The generated model.json and rewards.json files can be utilized in my portfolio repository (link) to create a captivating d3.js visualization.

Feel free to explore this repository and leverage the output files for a truly cool visualization experience!