Creating one hot vector from indices given as a tensor

1

I have a tensor of size 16 x 28 where 16 is batch size and 28 is sentence length. Every element of the sentence vectors are some index (0 to n). I want to create a 16 x 28 x n tensor where the vectors in 3rd dimension will be one hot encoding of the index which means I want to put 1 in the specified index and rest of the values will be zero. How can I do that using pytorch functionalities?

Right now, I am doing this with loop but I want to avoid looping!

2

If the one-hot vectors are used for retrieving word embeddings, just use Embedding layer instead.

3

No, this is not related to word embeddings at all.

4

You can use scatter for this purpose.

For example:

# your tensor of 16 x 28 dimensions, 
# where each element has some index (0 to n)
inp = torch.LongTensor(16, 28) % n    
inp_ = torch.unsqueeze(inp, 2)

one_hot = torch.FloatTensor(16, 28, n).zero_()
one_hot.scatter_(2, inp_, 1)

print(inp)
print(one_hot)
7 Likes