Crate onnxruntime_sys_ng
source ·Structs
- \brief Memory allocation interface
- \brief The C API
- \brief The helper interface to get the right version of OrtApi
- \brief CUDA Provider Options
- \brief MIGraphX Provider Options
- \brief OpenVINO Provider Options
- \brief ROCM Provider Options
- \brief TensorRT Provider Options
Constants
- < Error messages.
- < Fatal error messages (most severe).
- < Informational messages.
- < Verbose informational messages (least severe).
- < Warning messages.
- < Temporary CPU accessible memory allocated by non-CPU execution provider, i.e. CUDA_PINNED
- < Any CPU memory used by non-CPU execution provider
- < CPU accessible memory outputted by non-CPU execution provider, i.e. CUDA_PINNED
- < The default allocator for execution provider
Functions
- \brief The Onnxruntime library’s entry point to access the C API
- a64l⚠
- abs⚠
- atof⚠
- atoi⚠
- atol⚠
- bcmp⚠
- div⚠
- ecvt⚠
- exit⚠
- fcvt⚠
- ffs⚠
- ffsl⚠
- free⚠
- gcvt⚠
- l64a⚠
- labs⚠
- ldiv⚠
- rand⚠
Type Aliases
- \brief Graph optimization level
- Copied from TensorProto::DataType Currently, Ort doesn’t support complex64, complex128
- \brief Algorithm to use for cuDNN Convolution Op
- \brief Ort custom thread creation function
- \brief Custom thread join function
- \brief Language projection identifiers /see OrtApi::SetLanguageProjection
- \brief Logging severity levels
- \brief Memory types for allocated memory, execution provider specific types should be extended in each provider.
- \brief This mimics OrtDevice type constants so they can be returned in the API
- \brief Thread work loop function