Creates a model from a network function network-fn and a parameter set θ. The model is a function that returns an output tensor? given and input tensor.

Calculates the classification accuracy of the model as a real number between 0.0 and 1.0. ys is a tensor of one-hot tensors. Generates a tensor r by applying model to xs. The shapes of r and ys must be identical. Finally returns the number of positions in (argmax xs) and (argmax (model xs)) that are equal, divided by (tlen ys)