I am training a model with CTC, and I need to pad the input sequences for batches. However, the input length has to be at least 2*(output length)-1 because CTC has to output blanks between every output symbol. If I were to pad the output with pad tokens, do I need to pad the input with twice the amount of pad tokens? But this doesn't work because then the input size would be different for each sequence in a batch, depending on the output size.
Asked
Active
Viewed 48 times
