jamescasia/torch_rnn
Implementation of pytorch sequence models from scratch but allows defining output shape as parameter
Python
Issues
- 0
Add NaN input safety
#3 opened by jamescasia - 0
Sequence should be RNN with bidirectional, many to many/one to many, previous y when null input options
#6 opened by jamescasia - 0
Refactor to match Pytorch Module patterns
#5 opened by jamescasia - 0
Add support for LSTMs
#4 opened by jamescasia - 1
If RNN has output_dim >=2, and the seq_length < time_steps. The outputs will just be copied over and won't make sense.
#1 opened by jamescasia - 1