for j in range(1000):
MCTS(current)
data = torch.FloatTensor(StateToImg(current)).unsqueeze(0)
print(net(Variable(data)).data[0, 0])
best = current.child[0]
for c in current.child:
if c.n > best.n:
best = c
move2 = current.moves[current.child.index(best)]
GUI.Move(move2, current.pos[move2[0]])
current = best
data = torch.FloatTensor(StateToImg(current)).unsqueeze(0)
print(net(Variable(data)).data[0, 0])
if current.mandatory:
for move2 in current.moves:
GUI.SetBoard(True, *move2)
break
f = torch.load(PATH_PARAM)
net.load_state_dict(f)
net.eval()
for i in range(50):
MCTS(current)
GUI = Graphics(Callback)
GUI.Run()
