Suppose there are two torch tensors:

```
A=torch.arange(0,10)
B=torch.tensor([0,2,4,7,9])
```

First, I would like to produce a mask where B is missing values found in sequence A.

Then I would like to apply that mask to another tensor of size:

`C=torch.rand(5,2)`

So that for every index missing in B, zeros are filled on that row of C. The final output should be of size (10,2).

I could do it with a loop, but the size of these tensors are in the millions of rows. Is there a parallelized method to approach this problem?