How to assign values to specific blocks within a 2D tensor？

1

Hi! I want to realize this operation:
Here is a 2D tensor:
[[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]]
I want to set 1 to specific square blocks, such like this:
[[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,1,1,1,0,0,0,0,1,1,1,0,0,0],
[0,0,1,1,1,0,0,0,0,1,1,1,0,0,0],
[0,0,1,1,1,0,0,0,0,1,1,1,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,1,1,1,0,0,0,0,1,1,1,0,0,0],
[0,0,1,1,1,0,0,0,0,1,1,1,0,0,0],
[0,0,1,1,1,0,0,0,0,1,1,1,0,0,0],
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]]
I’ve tried this code before:

import torch 
import numpy as np 
import matplotlib.pyplot as plt

def random_patch_slice(h, w, p_size, sqrt_n):
  i_h = np.random.choice(range(0,h-p_size),size=sqrt_n,replace=False)
  i_w = np.random.choice(range(0,w-p_size),size=sqrt_n,replace=False)
  i_h = np.linspace(i_h, i_h + p_size-1,p_size,dtype=int,axis=1).flatten()
  i_w = np.linspace(i_w, i_w + p_size-1,p_size,dtype=int,axis=1).flatten()  
  return i_h, i_w

imgs = torch.zeros([10,10]) 
i_h,i_w = random_patch_slice(10, 10, 5, 2) 
imgs[i_h,i_w] = 1  
print(imgs)
plt.imshow(imgs)
plt.show()

But torch maps the indexes of the two dimensions one by one instead of the form of Cartesian product.
Then I tried this code:

import torch 
import numpy as np 
import matplotlib.pyplot as plt

def random_patch_slice(h, w, p_size, sqrt_n):
  i_h = np.random.choice(range(0,h-p_size),size=sqrt_n,replace=False)
  i_w = np.random.choice(range(0,w-p_size),size=sqrt_n,replace=False)
  i_h = np.linspace(i_h, i_h + p_size-1,p_size,dtype=int,axis=1).flatten()
  i_w = np.linspace(i_w, i_w + p_size-1,p_size,dtype=int,axis=1).flatten()  
  return i_h, i_w

imgs = torch.zeros([10,10]) 
i_h,i_w = random_patch_slice(10, 10, 5, 2) 
imgs[i_h][:,i_w] = 1   # modify here
print(imgs)
plt.imshow(imgs)
plt.show()

The value was just not assigned to the tensor. How can I realize this operation?
I would appreciate it if you can help me. :slightly_smiling_face: