I am having trouble training a quite complex dialog flow, and my testing doesn’t obtain a 100% accuracy. The flow is a bit complicated in that, in Stories, it has various checkpoints with around 4 possible inputs each. In the past I asked how to train this kind of dialog (which parameters to use during training), and you recommended modifying an additional parameter, which is the size of the RNN. In the past it worked, but the chatbot’s dialog flow has since become a bit more complex and now not even that can obtain 100% accuracy.
Is there an automated way to find the optimal training parameters given the Stories?