Improvements
#1
by
adamelliotfields
- opened
Reminder to circle back to this and make it usable. Also need to document the architecture in the README.
Essentially this is going to be a rewrite of Char-RNN-PyTorch in Keras 3. Should be able to reproduce their results after only 15 epochs.
The Space can be deployed as char-rnn and assuming everything works you can push a couple more models (additional characters for the space). For options, see the Lua torch-rnn.