Record Class MiniMaxShard

java.lang.Object
java.lang.Record
smile.llm.model.minimax.MiniMaxShard
Record Components:
rank - this rank, in [0, size).
size - number of ranks.
device - the rank's device; expert weights are allocated here.
public record MiniMaxShard(int rank, int size, Device device) extends Record
One rank of expert parallelism: this rank owns a contiguous range of the routed experts of every MoE layer and allocates their weights directly on device. Attention, dense FFNs, the router and the shared expert are replicated on every rank.

  • Constructor Summary

    Constructors
    Constructor
    Description
    MiniMaxShard(int rank, int size, Device device)
    Validates the rank.

  • Method Summary

    Modifier and Type
    Method
    Description
    Device
    device()
    Returns the value of the device record component.
    final boolean
    equals(Object o)
    Indicates whether some other object is "equal to" this one.
    int
    expertStart(int numExperts)
    Returns the first expert owned by this rank.
    final int
    hashCode()
    Returns a hash code value for this object.
    int
    localExperts(int numExperts)
    Returns the number of experts owned by this rank.
    int
    rank()
    Returns the value of the rank record component.
    static MiniMaxShard
    single()
    Returns the unsharded layout.
    int
    size()
    Returns the value of the size record component.
    final String
    toString()
    Returns a string representation of this record class.

    Methods inherited from class Object

    clone, finalize, getClass, notify, notifyAll, wait, wait, wait

  • Constructor Details

    • MiniMaxShard

      public MiniMaxShard(int rank, int size, Device device)
      Validates the rank.

  • Method Details

    • single

      public static MiniMaxShard single()
      Returns the unsharded layout.
      Returns:
      a single rank on the CPU (the model is moved with to(device)).

    • localExperts

      public int localExperts(int numExperts)
      Returns the number of experts owned by this rank.
      Parameters:
      numExperts - total routed experts.
      Returns:
      local expert count.

    • expertStart

      public int expertStart(int numExperts)
      Returns the first expert owned by this rank.
      Parameters:
      numExperts - total routed experts.
      Returns:
      global index of the first local expert.

    • toString

      public final String toString()
      Returns a string representation of this record class. The representation contains the name of the class, followed by the name and value of each of the record components.
      Specified by:
      toString in class Record
      Returns:
      a string representation of this object

    • hashCode

      public final int hashCode()
      Returns a hash code value for this object. The value is derived from the hash code of each of the record components.
      Specified by:
      hashCode in class Record
      Returns:
      a hash code value for this object

    • equals

      public final boolean equals(Object o)
      Indicates whether some other object is "equal to" this one. The objects are equal if the other object is of the same class and if all the record components are equal. Reference components are compared with Objects::equals(Object,Object); primitive components are compared with the compare method from their corresponding wrapper classes.
      Specified by:
      equals in class Record
      Parameters:
      o - the object with which to compare
      Returns:
      true if this object is the same as the o argument; false otherwise.

    • rank

      public int rank()
      Returns the value of the rank record component.
      Returns:
      the value of the rank record component

    • size

      public int size()
      Returns the value of the size record component.
      Returns:
      the value of the size record component

    • device

      public Device device()
      Returns the value of the device record component.
      Returns:
      the value of the device record component