Eden is a user on queer.cloud. You can follow them or interact with them if you have an account anywhere in the fediverse.
If you don't, you can sign up here.
QQ for anyone who does #PyTorch / #Tensorflow - how do you handle encoding words into one hot character strings?
Do you just pad the final tensor? Currently I can get it to work without minibatches, but the second I put it into a minibatch, it wants all the tensors to be the same shape, and obviously a word of length 6 is going to be a different tensor to a word of length 9.
Like I don't _hate_ having to pad it, it's just if I pad it then I have to do the onehot during data load rather than during the for loop, and if I do it in the for loop I can do it in one go for the batches.