Module tensor_core 

Unified tensor operations and device management Core Tensor Abstraction

This module provides the foundational tensor abstraction that unifies all tensor operations across CPU, GPU, and different data formats.

Uses Candle (HuggingFace’s Rust ML framework) as the computational backend.

Modules

ops_fn

Convenience functions for tensor operations

Structs

CandleStorage

Candle-backed tensor storage - the primary storage implementation

CpuStorage

CPU tensor storage implementation

CpuTensorOpsImpl

CPU implementation of tensor operations

GpuStorage

GPU tensor storage implementation (legacy - kept for compatibility)

GpuTensorOpsImpl

Basic GPU implementation of tensor operations

Tensor

Core tensor abstraction - the single source of truth for all tensor operations

TensorDispatcher

Tensor operation dispatcher - automatically selects CPU/GPU implementation

Enums

DataType

Data types supported by tensors

Device

Device abstraction

Traits

TensorOps

Core tensor operations trait - device-agnostic interface

TensorStorage

Tensor storage trait - abstracts how data is actually stored

Functions

ops

Get global tensor operations