ignition-monorepo

ignition-monorepo / backend-onnx/src / runInference

Function: runInference()

runInference(session, inputData, inputShape, inputName, outputName): Promise<Float32Array<ArrayBufferLike>>

Defined in: backend-onnx/src/runtime-universal.ts:55 

Runs a forward pass through an ONNX session.

Parameters

session

InferenceSession

Active InferenceSession

inputData

Float32Array

Flat Float32Array of the input tensor

inputShape

number[]

Shape of the input tensor (e.g. [1, 4])

inputName

string

ONNX graph input tensor name

outputName

string

ONNX graph output tensor name

Returns

Promise<Float32Array<ArrayBufferLike>>

Flat Float32Array of the output tensor (e.g. Q-values)