python numpy : 随机抽取二维矩阵中的多行或多列

import numpy as np
import torch

a = torch.rand((6,20,4))
a = a.cpu().numpy()

aaa = []
for i in range(0,6):
    aa = []
    for j,data in enumerate(a[i,:,-1]):
        if data > 0.5:
            b = a.tolist()
            aa.append(b[i][j])
        else: continue
    # 在每个子列表中随机抽取相同数量的值
    aa_np = np.array(aa)
    #   np二维数组的随机sample
    row_rand_aa = np.arange(aa_np.shape[0])
    np.random.shuffle(row_rand_aa)
    aa_sample_np = aa_np[row_rand_aa[0:6]]
    aa = aa_sample_np.tolist()
    aaa.append(aa)

aaa_np = np.array(aaa) # 子列表维度不相等无法转换成数组
print(aaa_np)

行随机抽取

row_rand_array = np.arange(array.shape[0])
np.random.shuffle(row_rand_array)
row_rand = array[row_rand_array[0:2]]

列随机抽取

col_rand_array = np.arange(array.shape[1])
np.random.shuffle(col_rand_array)
col_rand = array[col_rand_array[0:2]]