tflitecpu

Documentation

Overview

Package tflitecpu runs tflite model files on the host's CPU, as an implementation the ML model service.

Index

• func NewTFLiteCPUModel(ctx context.Context, params *TFLiteConfig, name resource.Name) (mlmodel.Service, error)
• type Model
  • func (m *Model) Infer(ctx context.Context, tensors ml.Tensors) (ml.Tensors, error)
  • func (m *Model) Metadata(ctx context.Context) (mlmodel.MLMetadata, error)
• type TFLiteConfig

Functions

func NewTFLiteCPUModel

func NewTFLiteCPUModel(ctx context.Context, params *TFLiteConfig, name resource.Name) (mlmodel.Service, error)

NewTFLiteCPUModel is a constructor that builds a tflite cpu implementation of the MLMS.

Types

type Model

type Model struct {
	resource.Named
	resource.AlwaysRebuild
	resource.TriviallyCloseable
	// contains filtered or unexported fields
}

Model is a struct that implements the TensorflowLite CPU implementation of the MLMS. It includes the configured parameters, model struct, and associated metadata.

func (*Model) Infer

func (m *Model) Infer(ctx context.Context, tensors ml.Tensors) (ml.Tensors, error)

Infer takes the input map and uses the inference package to return the result from the tflite cpu model as a map.

func (*Model) Metadata

func (m *Model) Metadata(ctx context.Context) (mlmodel.MLMetadata, error)

Metadata reads the metadata from your tflite cpu model into the metadata struct that we use for the mlmodel service.

type TFLiteConfig

type TFLiteConfig struct {
	resource.TriviallyValidateConfig
	// this should come from the attributes of the tflite_cpu instance of the MLMS
	ModelPath  string `json:"model_path"`
	NumThreads int    `json:"num_threads"`
	LabelPath  string `json:"label_path"`
}

TFLiteConfig contains the parameters specific to a tflite_cpu implementation of the MLMS (machine learning model service).