search

mlcommons / mobile_app_open

Mobile App Open
https://mlcommons.org/en/groups/inference-mobile/
Apache License 2.0
47 stars 23 forks source link

TensorFlow Lite Delegates Support Plan #897

Closed andyyeh75 closed 4 months ago

andyyeh75 commented 4 months ago

TensorFlow Lite Delegates https://github.com/mlcommons/mobile_app_open/issues/new

Delegates enable hardware acceleration of TensorFlow Lite models by leveraging on-device accelerators such as the GPU and Digital Signal Processor (DSP). (Warning: The NNAPI and Hexagon delegates are deprecated and no longer supported by TensorFlow Lite. )

Question: Just noticed the latest 4.0 still uses NNAPI for benchmark on current devices in the market. If the device can only support TFLite Stable Delegate, can later MLPerf Mobile support this modern ML interface?

andyyeh75 commented 4 months ago

Close this bug since believe TFLite has been supported.

v4.0

MLPerf Mobile 4.0 will also run on a host of other devices via our default path, which uses TensorFlow Lite on Android devices.