// the find
dennybritz/rnn-tutorial-rnnlm
Recurrent Neural Network Tutorial, Part 2 - Implementing a RNN in Python and Theano
A 2015 tutorial repo implementing a character/word-level language model RNN from scratch using Theano. It accompanies a blog post series on wildml.com and is aimed at people who want to understand backpropagation through time (BPTT) at the math level, not just call a library.
The implementation is didactic in the right way — rnn_theano.py is short enough to read in one sitting and maps directly to the math in the post. It includes a pre-trained model so you can run inference without waiting. The Reddit comments dataset is a genuinely good choice for a language model tutorial — realistic, messy, and large enough to matter.
Theano has been dead since 2017; nothing here runs without significant dependency archaeology, and the CUDA setup instructions target Ubuntu 14.10 with CUDA 7. There is no PyTorch or JAX equivalent, so the educational value is mostly theoretical now — you can read the math but you cannot easily run the code. The repo is also Part 2 of a series but stands alone poorly without the blog post; the notebook itself has minimal inline explanation.