Created by: dreamibor

Hi,

What's the input shape of the CTC decoder, is it (time_setps, num_classes)? The output from the CTC decoder is like "(nan, "ud rkn z ob elm c bi c vb wpx usr bg wr tok r weu ktt ty f ' pk ej stj w qe pd ayu gbs r jnl gc h bli po ggc h w j roi qk s e ech esf b wnm siba qa fu")" while it will be okay to use greedy decoder. I am using Python 3 and compile the CTC decoder for Python 3.5. Is that relevant?

Thanks in advance!