What accelerator is being used in CPU and GPU model android

[affereaflaw]

Description of issue (what needs changing)

please add more detail

Clear description

Hello, just out of curiosity, i am wondering what kind of accelerator is being use when I run CPU and GPU model like XNPACK/Vulkan/OpenCL/Others. And can we configure the delegate like former LiteRT?Thank you

Correct links

No response

Parameters defined

No response

Returns defined

No response

Raises listed and defined

No response

Usage example

No response

Request visuals, if applicable

No response

Submit a pull request?

No response

[kuaashish]

Hi @affereaflaw,

Currently, we believe the XNNPACK accelerator is used for the Android backend for both CPU and GPU. You can also see the former LiteRT (TFLite), which was available through the TFLite Interpreter API, but the LiteRT no longer supports it.

This seems like a valuable request that we can highlight with our team. Could you please raise a new feature request with complete details on the benefits of providing an option to choose the delegate? Based on future demand, we may work on this. In the meantime, if any possible workaround arises for using the accelerator, we will inform you accordingly.

Thank you!!

[kuaashish]

Apologies for the confusion. XNNPack is used for CPU, not GPU. However, LiteRT now supports this functionality, and you can find more details in their documentation. Our Task APIs use a different accelerator, while GenAI APIs support a separate accelerator. This is expected to be addressed soon, and you can find related documentation on this topic as it becomes available.

[affereaflaw]

@kuaashish thank you for the information, looking forward to try GenAI updates in accelerator