I think most of us have read this article although the article uses char-rnn, i’d like to use the current fastai library to repeat the results and see how far we’ve grown since we now can use awd-lstms with or without qrnn. Can someone guide me in that case ? i’m not extremely familiar with the fastai library
I too read Andrej Karpathy’s article and wanted to try out the fastai library for Char RNN. I wanted to replicate the Linux Source Code model described by Andrej. Obviously, transfer learning would not be of use, but would the workflow be to tokenize the source files by character and develop a language model that predicts the next character?