Struct ParallelConfig 

pub struct ParallelConfig {
    pub parallelism_type: ParallelismType,
    pub devices: Vec<Device>,
    pub tensor_parallel_size: usize,
    pub pipeline_parallel_size: usize,
    pub enable_memory_optimization: bool,
    pub communication_backend: String,
    pub max_chunk_size: usize,
    pub overlap_communication: bool,
}

Parallel execution configuration

Fields

parallelism_type: ParallelismType

Type of parallelism to use

devices: Vec<Device>

Devices to use for parallel execution

tensor_parallel_size: usize

Number of tensor parallel ranks

pipeline_parallel_size: usize

Number of pipeline parallel stages

enable_memory_optimization: bool

Whether to enable memory optimization

communication_backend: String

Communication backend (nccl, gloo, etc.)

max_chunk_size: usize

Maximum chunk size for all-reduce operations

overlap_communication: bool

Enable overlapping communication with computation

Implementations

impl ParallelConfig

pub fn single_gpu(device: Device) -> Self

Create config for single GPU

pub fn tensor_parallel(devices: Vec<Device>) -> Self

Create config for tensor parallelism

pub fn pipeline_parallel(devices: Vec<Device>) -> Self

Create config for pipeline parallelism

pub fn data_parallel(devices: Vec<Device>) -> Self

Create config for data parallelism

pub fn hybrid(devices: Vec<Device>, tp_size: usize, pp_size: usize) -> Self

Create hybrid config (tensor + pipeline)

pub fn world_size(&self) -> usize

Get the world size (total number of ranks)

pub fn is_parallel(&self) -> bool

Check if parallelism is enabled

pub fn device_for_rank(&self, rank: usize) -> Option<&Device>

Get device for a specific rank

pub fn tp_rank(&self, global_rank: usize) -> usize

Calculate tensor parallel rank from global rank

pub fn pp_rank(&self, global_rank: usize) -> usize

Calculate pipeline parallel rank from global rank

Trait Implementations

impl Clone for ParallelConfig

fn clone(&self) -> ParallelConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ParallelConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for ParallelConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for ParallelConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for ParallelConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.