Results 111 issues of Sean Moriarity
trafficstars

Right now the only way to construct a train step is using a loss function and an optimizer: ```elixir def train_step(model, loss, optimizer, opts \\ []) do ``` This is...

e.g. direct connection to Twilio

We have alignments, so we can truncate responses properly for interrupts

If you start talking while transcribing, it should get added to the queue and transcriptions should get appended and then processed for the next reply. No need to send to...