Welcome to my Reinforcement Learning (RL) project folder! 👋 This marks my very first attempt diving into the fascinating world of training agents to play games and solve environments. 🌱🔰 It's been quite the journey!
The main goal here was to learn the ropes of RL by tackling some classic environments:
- CartPole: Balancing that tricky pole! ⚖️
- LunarLander: Landing safely on the moon! 🌕🚀
- Atari Games: Braving the pixelated challenges of yesteryear! 👾 (Looks like I gathered a lot of ROMs! 💾)
- Racing Car: Vroom vroom! 🏎️💨
Oh boy, where do I start? 😂 This first attempt wasn't without its bumps:
- Hyperparameter Tuning Nightmare: Finding the right learning rate, discount factor, and network architecture felt like searching for a needle in a haystack! 🤯 So many trials, so many errors. 📈📉
- Reward Shaping Woes: Getting the agent to learn complex tasks, especially in environments with sparse rewards (like some Atari games), was tough! Sometimes the agent just wanted to spin in circles. 😵💫
- Environment Setup: Just getting Gym, Gymnasium, Stable Baselines3, and all the Atari dependencies to play nicely together took some serious debugging effort! 💻🔧🐍
- Understanding the Algorithms: Wrapping my head around PPO, DQN, and how they actually work required lots of reading and head-scratching. 📚🧠
Despite the hurdles, it's been an incredibly rewarding experience! Seeing an agent finally learn to play a game after hours of training is pure magic. ✨
.ipynbfiles: My notebooks containing the code, experiments, and maybe some frantic notes-to-self. 📝checkpoint/&RL Data/: Where the trained models and logs live (hopefully some successful ones!). 💾📊ROMS/: A collection of Atari ROMs for the ALE environment. 🕹️
Thanks for stopping by! Wish me luck on my next RL adventures! 🙏🚀