Struct TensorDecl 

pub struct TensorDecl {
    pub id: TensorId,
    pub name: String,
    pub role: TensorRole,
    pub elems: ElemShape,
    pub tiles: TileGrid,
    pub dtype: Precision,
    pub layout: Layout,
    pub space: TensorSpace,
    pub init: TensorInit,
    pub range: Option<(i64, i64)>,
    pub chip: Option<ChipId>,
}

Fields

id: TensorId

name: String

role: TensorRole

elems: ElemShape

tiles: TileGrid

dtype: Precision

layout: Layout

space: TensorSpace

init: TensorInit

range: Option<(i64, i64)>

Declared value range (inclusive), when known — feeds the numerics claim.

chip: Option<ChipId>

Which RDU holds this tensor, once a shard has split it. None on every unsharded program, and on the replicated tensors (activations, norms) of a sharded one.

Implementations

impl TensorDecl

pub fn tile_index(&self, row_t: usize, col_t: usize) -> usize

pub fn abs_bound(&self) -> Option<i64>

A bound on |value| from the fill or the declared range.

Trait Implementations

impl Clone for TensorDecl

fn clone(&self) -> TensorDecl

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TensorDecl

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for TensorDecl

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for TensorDecl

fn eq(&self, other: &TensorDecl) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for TensorDecl

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

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Eq for TensorDecl

impl StructuralPartialEq for TensorDecl