def train(nbSticks, nbGames, plotValueFunction, players, debugMode=False):
writer = Writer() if debugMode else None
game = Game(nbSticks, players, {
ConfigKey.TRAIN: True,
ConfigKey.DEBUG: debugMode
})
# stats for trained players
valueFunctionStats = {}
for p in (p for p in players if type(p) is TrainedPlayer):
valueFunctionStats[p] = dict((i, []) for i in range(1, nbSticks + 1))
print("\nTraining...")
for i in range(0, nbGames):
if writer:
writer.beginGame(i, game.players[0], game.players[1])
game.start() # play a game
if writer:
writer.endGame(game.players[0], game.players[1])
# update stats
for p in valueFunctionStats:
for s, v in p.v.items():
valueFunctionStats[p][s].append(v)
# decrease exploration over time
if i % 10 == 0:
for p in players:
p.epsilon = max(p.epsilon * 0.996, 0.05)
game.reset()
if writer:
writer.close()
for p in (p for p in players if type(p) is TrainedPlayer):
print(f'\n=== Value function of {p}: ===')
PrettyPrinter().pprint(p.v)
if plotValueFunction:
plot(valueFunctionStats)
