A Mulit-Agent Reiforcement Learning model to play chess

Based on code from PettingZoo and Tianshou

Pre-Trained Agents

You can watch a number of pretrained agents play against each other.

Note: mix_train.pth and mix_train_adv.pth can only play against each other due to architecture differences `

Random.pth

Agent trained for 60 epochs against player that plays random moves

self_play.pth

Agent trained against itself for 60 epochs

mixed_train.pth

Agent trained with both self play and against random players for 75 epochs

mix_train.pth

Same as mixed_train.pth but model has two layers with 256 neurons. Trained for 60 epochs

mix_train_adv.pth

Same as mix_train.pth but trained for 100 epochs