com.google.android.gms.tflite.acceleration

  • The content describes interfaces and classes for configuring and validating TensorFlow Lite acceleration on Android.

  • Key classes include AccelerationConfig for different acceleration types like CPU and GPU, AccelerationService for the API, and BenchmarkResult for performance data.

  • Validation configurations, such as CustomValidationConfig and EmbeddedValidationConfig, are available for benchmarking model accuracy.

  • Various enums define options for accelerators, benchmark stages, and GPU settings like backend, priority, and usage.

Interfaces

CustomValidationConfig.AccuracyValidator Defines custom accuracy validation rule. 

Classes

AccelerationConfig Abstract class that represents all types of acceleration configs. 
AccelerationConfigFactory Factory class to create concrete AccelerationConfig. 
AccelerationService Acceleration Service API  
BenchmarkError Data class that contains BenchmarkError. 
BenchmarkResult Data class that contains BenchmarkResult. 
BenchmarkResult.BenchmarkMetric Accuracy metrics. 
BenchmarkResult.InferenceOutput Model output of one tensor. 
CpuAccelerationConfig Concrete class that represents CPU acceleration configs. 
CpuAccelerationConfig.Builder Builder class. 
CustomValidationConfig Provides custom validation data for benchmarking. 
CustomValidationConfig.Builder Builder class. 
EmbeddedValidationConfig Use this class when the validation data is already embedded in the model. 
EmbeddedValidationConfig.Builder Builder class. 
GpuAccelerationConfig Concrete class that represents GPU acceleration configs. 
GpuAccelerationConfig.Builder Builder class. 
Model Model information. 
Model.Builder Builder class. 
Model.ModelLocation Where to load the model from. 
ValidatedAccelerationConfigResult Acceleration config and validation result. 
ValidationConfig Abstract class that defines validation configs. 

Enums

AccelerationConfig.AcceleratorType Accelerator to use. 
BenchmarkError.BenchmarkStage When during benchmark execution an error occurred. 
GpuAccelerationConfig.GpuBackend Which GPU backend to select. 
GpuAccelerationConfig.GpuInferencePriority Relative priorities given by the GPU delegate to different client needs. 
GpuAccelerationConfig.GpuInferenceUsage GPU inference preference for initialization time vs.