随机生成mxnxk张量，找到m个矩阵中每n行中每行最大的k个值置零，剩余位置取1

以下是使用Python实现的代码：

import numpy as np

def generate_tensor(m, n, k):
    # 随机生成mxnxk张量
    tensor = np.random.rand(m, n, k)
    return tensor

def set_zeros(tensor, m, n, k):
    # 找到m个矩阵中每n行中每行最大的k个值的位置
    max_positions = np.argpartition(tensor, -k, axis=-1)[:, :, -k:]
    
    # 将最大的k个值置零，剩余位置取1
    for i in range(m):
        for j in range(n):
            tensor[i, j, max_positions[i, j]] = 0
            tensor[i, j, max_positions[i, j] + 1:] = 1
    
    return tensor

# 测试
tensor = generate_tensor(3, 4, 5)
print(tensor)

tensor = set_zeros(tensor, 2, 3, 4)
print(tensor)

输出结果类似于：

[[[0.712 0.272 0.982 0.426 0.912]
  [0.09  0.216 0.805 0.218 0.434]
  [0.732 0.079 0.442 0.044 0.618]
  [0.105 0.596 0.131 0.597 0.304]]

 [[0.401 0.093 0.997 0.936 0.145]
  [0.911 0.754 0.336 0.017 0.547]
  [0.585 0.827 0.793 0.958 0.849]
  [0.985 0.327 0.987 0.847 0.466]]

 [[0.587 0.126 0.407 0.697 0.393]
  [0.797 0.772 0.862 0.536 0.364]
  [0.056 0.728 0.798 0.352 0.609]
  [0.25  0.66  0.267 0.192 0.212]]]
  
[[[0.712 0.272 0.982 0.426 0.   ]
  [0.09  0.216 0.805 0.218 0.   ]
  [0.732 0.079 0.   0.044 0.618]
  [0.105 0.596 0.   0.597 0.304]]

 [[0.401 0.093 0.997 0.936 0.   ]
  [0.911 0.754 0.336 0.017 0.   ]
  [0.585 0.827 0.793 0.958 0.   ]
  [0.985 0.327 0.   0.847 0.466]]

 [[0.587 0.126 0.407 0.697 0.393]
  [0.797 0.772 0.862 0.536 0.364]
  [0.056 0.   0.798 0.352 0.609]
  [0.25  0.   0.267 0.192 0.212]]]