🃏 Blackjack Agent Project

Project Overview

This project is a Streamlit-based Blackjack Agent that uses PyTorch to load a trained model and make decisions in a simplified Blackjack game. The app allows users to input the game state (player's hand sum, dealer's visible card, usable ace) and get the agent's recommended action (hit or stick).

Features

• Interactive UI built with Streamlit
• Loads a trained PyTorch model
• Accepts user input for game state
• Outputs the agent's recommended action
• Lightweight and easy to deploy

Installation

1. Clone the repository:

   git clone https://github.com/BrianP490/BlackJack-Agent.git
   cd BlackJack-Agent

2. Create and Activate the Conda Environment for local app hosting: This project uses a Conda environment for managing dependencies, including PyTorch. The environment configuration is provided in dev_requirements.yaml

   • Create the environment: Ensure you are in the root directory of the cloned repository (BlackJack-Agent/).

   conda env create -f dev_requirements.yaml

   • Activate the environment:

   conda activate GRPO_env

   Your terminal prompt should change to (GRPO_env) indicating the environment is active.

   
   If running ipynb file on VS code: Select install ikernel when popup appears when first activating the env, select the GRPO_env for the kernel, and run the cells.

3. (OPTIONAL) Train a New Model: You can also run the training script "./app/scripts/train.py":

   # Use the -h option for running options:
   python ./app/scripts/train.py -h

   You will have to change line 10's variable "weights_file" in "./app/main.py" to use the correct path to your trained model weights. This is if you chose to save you model in another location besides the default path "./app/model_weights/blackjack_policy_model.pth"

4. Run the Streamlit Application: Once your Conda environment is active, you can launch the Streamlit application.

   • Navigate to the correct directory (if not already there): Make sure your current working directory in the terminal is BlackJack-Agent/ (the root of the project where app/ resides).

   # If you're not in the BlackJack-Agent directory, navigate there:
   # cd C:\Users\brian\DESKTOP\Dev\GIT_CLONES\projects\GRPO_Project_Iteration\BlackJack

   • Launch the Streamlit app:

   streamlit run app/main.py
   

   Enjoy!

Launching Using Streamlit.io

You can also launch this with the streamlit online app deployment on https://streamlit.io/, just make sure to follow the steps to create a new repository with this clone and make sure that "requirements.txt" is at the root directory.