问pytorch中调用cuda计算，怎么将多个任务打包一次性发送给GPU进行计算？

import torch

length = 7000
num = 500000
device = "cuda:0"
a1 = torch.randint(0, 2, (num, length), dtype=torch.bool, device=device)
a2 = torch.randint(0, 2, (num, length), dtype=torch.bool, device=device)
b1 = torch.zeros(num, dtype=torch.bool, device=device)
b2 = torch.zeros(num, dtype=torch.bool, device=device)
c1 = torch.zeros(num, dtype=torch.float32, device=device)
c2 = torch.zeros(num, dtype=torch.float32, device=device)
d = torch.zeros(num, dtype=torch.int32, device=device)
e = torch.randint(0, 101, (length,), dtype=torch.float32, device=device)
f = torch.ones(num, dtype=torch.bool, device=device)

for i in range(length):
    sum_data = e[i]
    a1_ = a1[:, i].flatten()
    a2_ = a2[:, i].flatten()

    b1[a1_] = True
    b2[a2_] = True

    c1[b1] += sum_data
    c2[b2] += c1[b2]
    d[b2] += 1
    b1[b2] = False
    b2[f] = False