dsl

Type Members

1. case class Attributes(lookup: Map[String, Value]) extends Product with Serializable

   

   Class that holds a map of string to ml.combust.bundle.dsl.Value objects.

   Class that holds a map of string to ml.combust.bundle.dsl.Value objects.

   Can only have one attribute with a given name at a time.

   lookup

   map of attribute name to ml.combust.bundle.dsl.Value object

2. case class Bundle[Transformer <: AnyRef](info: BundleInfo, root: Transformer) extends Product with Serializable

   

   Root object for serializing Bundle.ML pipelines and graphs.

   Root object for serializing Bundle.ML pipelines and graphs.

   info

   info data for the bundle

   root

   root transformer node

3. case class BundleInfo(uid: UUID, name: String, format: SerializationFormat, version: String, timestamp: String, meta: Option[bundle.Attributes]) extends Product with Serializable

   

   Information data for a bundle.

   Information data for a bundle.

   uid

   uid for the bundle

   name

   name of the bundle

   format

   serialization format of the Bundle

   version

   Bundle.ML version used for serializing

   timestamp

   LocalDateTime when the model was created

4. trait HasAttributes[T] extends AnyRef

   

   This trait provides easy access to reading/writing attributes to objects that contain an Attributes.

   This trait provides easy access to reading/writing attributes to objects that contain an Attributes.

   T

   class type that contains the attribute list

5. case class Model(op: String, attributes: Attributes = Attributes()) extends HasAttributes[Model] with Product with Serializable

   

   Class that encodes all information need to serialize or deserialize a machine learning model.

   Class that encodes all information need to serialize or deserialize a machine learning model.

   Models encode things like coefficients for linear regressions, labels for string indexers, sizes for one hot encoders, decision trees, and any other data needed to serialize and deserialize ML models and feature builders.

   op

   op name for the model

   attributes

   optional list of attributes for the model

6. case class Node(name: String, shape: NodeShape) extends Product with Serializable

   

   Class for storing a node in the Bundle.ML graph.

   Class for storing a node in the Bundle.ML graph.

   Bundle.ML is composed of a set of Node objects, each with their own inputs and outputs defined by their Node#shape.

   Every Node needs a unique identifier within a Bundle.ML graph. This is the Node#name.

   name

   unique identifier for the node

   shape

   shape of the node

7. case class NodeShape extends Product with Serializable

   

   Class for holding the input fields and output fields of a Node.

   Class for holding the input fields and output fields of a Node. The shape also holds information for connecting the input/output fields to the underlying ML model.

   A NodeShape contains input and output sockets. Sockets map field data to certain functionality within a Model. For instance, say we want to run a "label" field through a string indexer and have the result output to the field "label_name". We could wire up the node like so:

   scala> import ml.bundle.dsl._
   scala> Shape().withInput("label", "input"). // connect the "label" field to the model input
            withOutput("label_name", "output") // connect the model output to the "label_name" field

   Or more concisely:

   scala> import ml.bundle.dsl._
   scala> Shape().withStandardIO("label", "label_name") // shorthand for the above code

8. case class Value(value: bundle.Value) extends Product with Serializable

   

   This class is used to wrap Scala objects for later serialization into Bundle.ML

   This class is used to wrap Scala objects for later serialization into Bundle.ML

   value

   bundle protobuf value