RuntimeError: bool value of Variable objects containing non-empty torch.cuda.FloatTensor is ambiguous

[Zhanchenhaha]

Traceback (most recent call last):
File "main.py", line 150, in
val_loss, val_rae, val_corr = evaluate(Data, Data.valid[0], Data.valid[1], model, evaluateL2, evaluateL1, args.batch_size);
File "main.py", line 24, in evaluate
if not predict:
File "/usr/lib64/python2.7/site-packages/torch/autograd/variable.py", line 125, in bool
torch.typename(self.data) + " is ambiguous")
how to deal it