Class Tensor

Tensor( n-dimnesional array).

Hierarchy (View Summary)

TVMObject
- Tensor

Index

Constructors

Properties

device dtype isView ndim shape

Methods

copyFrom copyFromRawBytes dispose getDataPtr getHandle toArray toRawBytes typeIndex typeKey view

Constructors

constructor

new Tensor(
    handle: number,
    lib: FFILibrary,
    ctx: RuntimeContext,
    isView: boolean,
): Tensor
Parameters
- handle: number
- lib: FFILibrary
- ctx: RuntimeContext
- isView: boolean
Returns Tensor
Overrides TVMObject.constructor
- Defined in runtime.ts:518

Properties

device

device: DLDevice

Device of the array.

dtype

dtype: string

Data type of the array.

isView

isView: boolean

Whether it is a temporary view that can become invalid after the call.

ndim

ndim: number

Number of dimensions.

shape

shape: number[]

Shape of the array.

Methods

copyFrom

copyFrom(
    data:
        | number[]
        | Uint8Array<ArrayBufferLike>
        | Int32Array<ArrayBufferLike>
        | Float32Array<ArrayBufferLike>
        | Float64Array<ArrayBufferLike>
        | Tensor
        | Int8Array<ArrayBufferLike>
        | Uint8ClampedArray<ArrayBufferLike>,
): this
Copy data from another Tensor or javascript array. The number of elements must match.
Parameters
- data:
      | number[]
      | Uint8Array<ArrayBufferLike>
      | Int32Array<ArrayBufferLike>
      | Float32Array<ArrayBufferLike>
      | Float64Array<ArrayBufferLike>
      | Tensor
      | Int8Array<ArrayBufferLike>
      | Uint8ClampedArray<ArrayBufferLike>
  The source data array.
Returns this
this
- Defined in runtime.ts:605

copyFromRawBytes

copyFromRawBytes(data: Uint8Array): this
Copy data from raw bytes.
Parameters
- data: Uint8Array
  Uint8Array of bytes.
Returns this
this
- Defined in runtime.ts:648

dispose

dispose(): void
Dispose the internal resource This function can be called multiple times, only the first call will take effect.

Returns void
Inherited from TVMObject.dispose
- Defined in runtime.ts:449

getDataPtr

getDataPtr(): number
Get dataPtr of NDarray

Returns number
The handle.
- Defined in runtime.ts:591

getHandle

getHandle(requireNotNull?: boolean): number
Get handle of module, check it is not null.
Parameters
- requireNotNull: boolean = true
  require handle is not null.
Returns number
The handle.
Inherited from TVMObject.getHandle
- Defined in runtime.ts:464

toArray

toArray(): | Uint8Array<ArrayBufferLike>
| Int32Array<ArrayBufferLike>
| Float32Array<ArrayBufferLike>
| Float64Array<ArrayBufferLike>
| Int8Array<ArrayBufferLike>
Return a TypedArray copy of the Tensor, the specific type depends on the dtype of the Tensor.

Returns
    | Uint8Array<ArrayBufferLike>
    | Int32Array<ArrayBufferLike>
    | Float32Array<ArrayBufferLike>
    | Float64Array<ArrayBufferLike>
    | Int8Array<ArrayBufferLike>
The result array.
- Defined in runtime.ts:681

toRawBytes

toRawBytes(): Uint8Array
Return a copied Uint8Array of the raw bytes in the Tensor.

Returns Uint8Array
The result array.
- Defined in runtime.ts:669

typeIndex

typeIndex(): number
get the type index of the object

Returns number
Inherited from TVMObject.typeIndex
- Defined in runtime.ts:472

typeKey

typeKey(): string
get the type key of the object

Returns string
Inherited from TVMObject.typeKey
- Defined in runtime.ts:480

view

view(shape: number[], dtype?: string): Tensor
Create a view of the array.
Parameters
- shape: number[]
  The shape of the view.
- Optionaldtype: string
  The data type of the new array.
Returns Tensor
The new sliced ndarray.
- Defined in runtime.ts:574