This repository has been archived by the owner on Jan 1, 2024. It is now read-only.
Optimal result for first overnight training
- epoch 1k
- avg score 22.5
- agent config:
{
'adamConfig': {
'stepSize': 1e-3,
'betaM': 0.9,
'betaV': 0.999,
'epsilon': 0.001
},
'nnConfig': {
'stateCount': 12,
'hiddenUnitCount': [128, 64, 16],
'actionCount': 4,
},
'rbConfig': {
'rbSize': 50000,
'batchSize': 8,
'replayUpdatePerStep': 4,
},
'gamma': 0.95,
'tau': 0.001
}