Enum Qnn_TensorType_t

Enum Documentation

enum Qnn_TensorType_t

An enum to specify the tensor type, application accessible or native to QNN.

Values:

enumerator QNN_TENSOR_TYPE_APP_WRITE = 0

Client application writeable tensor.

enumerator QNN_TENSOR_TYPE_APP_READ = 1

Client application readable tensor.

enumerator QNN_TENSOR_TYPE_APP_READWRITE = 2

Tensor that can both be read and written by an application. Used in scenarios that may include supplying an output tensor from one graph as the input to another graph.

enumerator QNN_TENSOR_TYPE_NATIVE = 3

Tensor native to a graph which may be optimized by a backend and are not accessible by a client.

enumerator QNN_TENSOR_TYPE_STATIC = 4

Static data which doesn’t change during execution and may be optimized by a backend. Since the data cannot change, static tensors cannot have dynamic dimensions.

enumerator QNN_TENSOR_TYPE_NULL = 5

Tensor type NULL which can be used to represent optional tensors. Other Qnn_Tensor_t metadata is ignored.

enumerator QNN_TENSOR_TYPE_UPDATEABLE_STATIC = 6

Tensor containing static data whose content or quantization encodings may be modified by a client after tensor creation.

enumerator QNN_TENSOR_TYPE_UPDATEABLE_NATIVE = 7

Tensor native to a graph whose quantization encodings may be modified by a client after tensor creation.

enumerator QNN_TENSOR_TYPE_UPDATEABLE_APP_WRITE = 8

Application writable tensor whose quantization encodings may be modified by a client after tensor creation.

enumerator QNN_TENSOR_TYPE_UPDATEABLE_APP_READ = 9

Application readable tensor whose quantization encodings may be modified by a client after tensor creation.

enumerator QNN_TENSOR_TYPE_UPDATEABLE_APP_READWRITE = 10

Application readable/writable tensor whose quantization encodings may be modified by a client after tensor creation.

enumerator QNN_TENSOR_TYPE_OPTIONAL_APP_WRITE = 11

Tensor type OPTIONAL_APP_WRITE represents an application writable (input) tensor that may be excluded from inferences

enumerator QNN_TENSOR_TYPE_OPTIONAL_APP_READ = 12

Tensor type OPTIONAL_APP_READ represents an application readable (output) tensor that may be excluded from inferences.

enumerator QNN_TENSOR_TYPE_OPTIONAL_APP_READWRITE = 13

Tensor type OPTIONAL_APP_READ_WRITE represents an application readable (output) or writable (input) tensor that may be excluded from inferences.

enumerator QNN_TENSOR_TYPE_UNDEFINED = 0x7FFFFFFF