Module distributed 

Re-exports

pub use ring_allreduce::CollectiveTransport;

pub use ring_allreduce::LocalTransport;

pub use ring_allreduce::ReduceOp;

pub use ring_allreduce::RingAllReduce;

pub use pipeline_parallel::OpKind;

pub use pipeline_parallel::PipelineConfig;

pub use pipeline_parallel::PipelineExecution;

pub use pipeline_parallel::PipelineMetrics;

pub use pipeline_parallel::PipelineOp;

pub use pipeline_parallel::PipelineSchedule;

pub use pipeline_parallel::PipelineScheduler;

pub use pipeline_parallel::StageCost;

pub use pipeline_parallel::StagePartitioner;

pub use pipeline_parallel::StageRange;

pub use elastic::ElasticConfig;

pub use elastic::ElasticCoordinator;

pub use elastic::EpochSnapshot;

pub use elastic::MembershipEvent;

pub use elastic::RendezvousState;

pub use elastic::ShardRange;

pub use elastic::WorkerShards;

pub use elastic::WorkerState;

pub use fedprox::*;

Modules

elastic

fedprox

pipeline_parallel

ring_allreduce

Structs

CommunicationResult

Result of a communication round

CompressedGradient

Compressed gradient representation

CompressionMetadata

Compression metadata

CompressionStats

Compression statistics

DistributedCoordinator

Distributed training coordinator

GradientCompressor

Gradient compression engine

ParameterAverager

Distributed parameter averager

ParameterServer

Synchronous parameter server for distributed training

TrainingStats

Training statistics for distributed training

Enums

AveragingStrategy

Parameter averaging strategies for distributed training

CompressionStrategy

Gradient compression strategies for communication optimization