Resumable model training with a status summary.
One iteration of learning, returning some summary of the computation
Hard maximum number of iterations of learning.
Target for saving the final model (and periodic intermediate models if outputIterations is defined).
Saves the state of the model to the given folder. Mandatory if output is not None.
Loads the state of the model from the given folder. Mandatory if output is not None.
Some (human-readable) description of the source of the model and data. For checking if the right model is stored on disk for resumed training.
If defined, saves the intermediate state of the model (and its history) every outputIterations iterations (e.g. every 50 iterations if outputIterations=Some(50)).
Test if the model has converged from a list of iteration numbers and summaries. The length of the list is maxHistory if maxHistory is defined, else the length of the history list is unbounded.
Maximum length of history provided to convergence.
Where status messages go.