Module dlpack 

DLPack protocol for zero-copy tensor exchange.

Implements __dlpack__ and __dlpack_device__ for arrays managed by this crate. DLPack is a standard open-source ABI used by PyTorch, JAX, TensorFlow, and other frameworks to exchange tensors without copying.

Reference: https://dmlc.github.io/dlpack/latest/

Structs

DLDataType

DLPack data-type descriptor (ABI-compatible with the DLPack spec).

DLDevice

DLPack device descriptor.

DLManagedTensor

Managed DLPack tensor with associated deleter callback.

DLPackCapsule

Python class wrapping a DLPack-compatible tensor.

DLTensor

The core DLPack tensor structure (ABI-compatible).

DLTensorInfo

Structured information extracted from a validated DLTensor.

Enums

DLDataTypeCode

Element type codes used in DLPack DLDataType.code.

DLDeviceType

Device type codes used by DLPack.

DlpackError

Errors produced by DLPack validation and conversion utilities.

JaxDeviceType

Known device types as reported by JAX DLPack tensors.

Constants

DL_DEVICE_TYPE_TPU

Device type constant for JAX TPU tensors.

Functions

array_from_dlpack_f32^⚠

Generic DLPack array construction — accepts f32 tensors from any framework.

array_from_dlpack_f64^⚠

Generic DLPack array construction — accepts f64 tensors from any framework.

check_tensor_contiguous

Check that a DLTensor’s strides represent a C-order (row-major) contiguous layout.

dlarray_from_torch_f32^⚠

Convert a raw DLTensor pointer (from a PyTorch DLPack capsule) to an ndarray dynamic array view of f32 elements.

dlarray_from_torch_f64^⚠

Convert a raw DLTensor pointer (from a PyTorch DLPack capsule) to an ndarray dynamic array view of f64 elements.

dlpack_from_slice

Create a DLTensor that borrows data and shape slices.

dlpack_to_vec_f64

Extract a Vec<f64> from a CPU float64 DLTensor.

jax_device_type

Classify the device reported by a DLTensor as a JaxDeviceType.

register_dlpack_module

Register DLPack classes into a PyO3 module.

validate_dlpack_tensor

Validate a DLTensor and extract structured metadata.

validate_jax_dlpack_tensor

Validate that a DLTensor is compatible with JAX DLPack interop.

validate_torch_dlpack_tensor

Validate that a DLTensor is compatible with PyTorch DLPack interop.