Constants

The following constants are defined in dpctl.tensor:

dpctl.tensor.DLDeviceType = <enum 'DLDeviceType'>

An enum.IntEnum for the types of DLDevices supported by the DLPack protocol.

kDLCPU:

CPU (host) device

kDLCUDA:

CUDA GPU device

kDLCUDAHost:

Pinned CUDA CPU memory by cudaMallocHost

kDLOpenCL:

OpenCL device

kDLVulkan:

Vulkan buffer

kDLMetal:

Metal for Apple GPU

kDLVPI:

Verilog simulator buffer

kDLROCM:

ROCm GPU device

kDLROCMHost:

Pinned ROCm CPU memory allocated by hipMallocHost

kDLExtDev:

Reserved extension device type used to test new devices

kDLCUDAManaged:

CUDA managed/unified memory allocated by cudaMallocManaged

kDLOneAPI:

Unified shared memory allocated on a oneAPI non-partitioned device

kDLWebGPU:

Device support for WebGPU standard

kDLHexagon:

Qualcomm Hexagon DSP

kDLMAIA:

Microsoft MAIA device

dpctl.tensor.e

float:

IEEE 754 floating-point representation of Euler’s constant.

dpctl.tensor.inf

float:

IEEE 754 floating-point representation of (positive) infinity.

dpctl.tensor.nan

float:

IEEE 754 floating-point representation of Not a Number (NaN).

dpctl.tensor.newaxis

NoneType:

Alias for None which is useful for indexing.

dpctl.tensor.pi

float:

IEEE 754 floating-point representation of the mathematical constant π.