can't convert a given np.ndarray to a tensor

Question

source_label_ce = Variable(torch.from_numpy(ids_cvt[source_label.argmax(-1)]).long(), requires_grad=False).cuda()

CE_loss = CE(source_seen_output * self.cross_entropy_lambda, source_label_ce)

RuntimeError: can't convert a given np.ndarray to a tensor - it has an invalid type. The only supported types are: double, float, int64, int32, and uint8. pytorch报错怎么改？

Answer 1

由于读入的numpy数组里的元素是object类型，无法将这种类型转换成tensor。

所以，将numpy数组进行强制类型转换成float类型（或者任何pytorch支持的类型：float64, float32, float16, int64, int32, int16, int8, uint8, and bool）即可。
如

trainx = trainx.astype(float)