The data files for the Word example are located in data/word/
and the files for the Line example in data/line
.
Each of these directories contains:
rnnOutput.csv
: output of RNN layer (softmax not yet applied), which contains 32 or 100 time-steps and 80 label scores per time-step.corpus.txt
: the text from which the language model is generated.img.png
: the input image of the neural network. It is contained as an illustration, however, the decoding algorithms do not use it.