pytorch中tensor与其他数据结构的转化(list, array)

本文记录cpu张量、cuda张量、list和array之间的转换关系。 

import torch
import numpy as np
# list -> tensor(cpu)
l0 = [1, 2, 3]
t = torch.Tensor(l0)# tensor(cpu) -> numpy -> list
a = t.numpy()
l1 = t.numpy().tolist()# list -> numpy
a0 = np.array(l0)# numpy -> tensor(cpu)
t1 = torch.from_numpy(a0)# tensor(cpu) -> tensor(cuda)
tc = t1.cuda()# tensor(cuda) -> list
l2 = tc.cpu().numpy().tolist()