DataParallel models can do assignments in parallel, updating model parameters locally, aggregating model parameters upwards, e.
Inference parameters for gibbs sampling.
Models that can update their parameters from looking at one example at a time.
Models that support a log probability estimate.
Generic interface for trainable models with both model state and data state.
A Modeler holds a model and a set of data items, allowing certain operations to be performed in aggregate on that data.
Mechanism for composing representation checks.
Single-threaded modeler that holds data items in an internal ListBuffer.
A model companion that provides an implicit FileSerialization backing for a model where the parameters and state are both ReadWriteable.
A unified collection-like view of a set of shards, each of which is IterableLike.
Runs data parallel models as multiple threads on a single machine.
Inference in Gibbs Samplers based on updating a given result distribution using samples drawn from a chain.
Static method for resumable training of a model.