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.