kradonneoh

At least for the numba debug issue, I assume you're running this in the Google Colab notebook. It's not running infinitely, the stdout on the colab is just so slow...

+1 I can reverse engineer the forward function, but it'd be nice if there was an easy function call I'm missing Edit: ended up reverse engineering it anyways :-) You...

Ah, I must've misread the code / the paper -- when we call `setup_rnn`, the kernel state matrices are initialized for step calls only, but not for regular forward calls?...

While I have you, I'm attempting to understand the output of the Sashimi forward pass and how I'd do a simple cross entropy loss on the predictions. If I'm reading...

Sweet, that lines up with my intuition. One last question for you: when trying to generate _conditionally_, my thought was to convert to RNN, initialize default state, then feed the...

Thank you, appreciate the quick response times.