Crate tenferro_tensor_core 

Lightweight host tensor data model and metadata-only views.

tenferro-tensor-core owns backend-independent tensor metadata and host-resident contiguous tensor storage. It does not own execution backends, backend buffers, GPU handles, provider selection, or materializing kernels. Runtime/backend-capable TypedTensor<T, R> lives in tenferro-tensor. This crate exposes rank/layout metadata plus host-only tensor adapters.

Examples

use tenferro_tensor_core::{HostTensor, Rank, SliceSpec, TensorLayout};

let tensor = HostTensor::from_vec_col_major(vec![2, 3], vec![1.0_f64, 2.0, 3.0, 4.0, 5.0, 6.0])?;
let view = tensor
    .as_view()
    .slice_view(&[
        SliceSpec { start: 0, end: 2, step: 1 },
        SliceSpec { start: 1, end: 3, step: 1 },
    ])?;

assert_eq!(view.shape(), &[2, 2]);
assert_eq!(view.as_slice()?, &[3.0, 4.0, 5.0, 6.0]);

let layout = TensorLayout::<Rank<2>>::compact([2, 3])?;
let transposed = layout.transpose_view([1, 0])?;
assert_eq!(transposed.shape(), &[3, 2]);

Structs

DynRank

Dynamic tensor rank marker.

HostTensor

Owned contiguous host tensor in column-major order.

HostTensorView

Borrowed host tensor view with shape, strides, and offset metadata.

Rank

Static tensor rank marker.

SliceSpec

Explicit slice descriptor.

TensorLayout

Storage-neutral tensor layout metadata.

Enums

DType

Runtime scalar dtype tag.

Error

Data-model validation errors.

Tensor

Dynamic owned host tensor over the supported dtype set.

TensorRef

Core-neutral tensor input reference.

TensorView

Dynamic borrowed host tensor view.

Traits

TensorRank

Rank contract for tensor metadata shapes and strides.

TensorScalar

Sealed trait for scalar types supported by the core tensor data model.

Functions

col_major_strides

Return compact column-major strides for a shape.

Type Aliases

Result

Result type for tensor data-model operations.

ShapeVec

Small tensor shape vector with inline capacity for common dynamic ranks.

StrideVec

Small tensor stride vector with signed element strides.