A deep Q-learning neural network for Gymnasium's blackjack environment

Model: 4 linear layer model

Loss: Mean squared error

Trained on 1000 games then visually plays 1000 games

Note: Here instead of having a dedicated target and current network I just have one for simplicity.