TMS320C6748: How to Port or Install Machine Learning?

Zhi Sheng

Part Number: TMS320C6748

Hi,

I am using the TMS320C6748 LDCK with XDS110 Debug probe. I am perform A sampling of DFT of 1000points with timing shifting. I am looking at the tensorflow lite library but i am not sure how to port the library to the board.

over 4 years ago

0 Jianzhong Xu over 4 years ago

TI__Guru 60890 points

Hi,

I don't think that is possible, because we don't have a supported TensorFlow Lite platform (please refer here) running on the C6748 DSP. I would recommend a few alternative options:

1. OMAP-L138 which has Cortex-A9 core and C674x DSP. We have Linux Processor-SDK running on the Cortex-A9 core and you should be able to cross-compile TFLite for it.

2. We have built-in TFLite support for Sitara devices.

3. You may be able to port TensorFlow Lite for Microcontrollers to C6748 DSP. I have not tried this myself but you may want to check it out and see if it is doable and meets your needs.

Regards,

Jianzhong

0 Zhi Sheng over 4 years ago in reply to Jianzhong Xu

Prodigy 130 points

I may have referring to your options 2. But I am not sure how to activate it. After installing the SDK, do I just include the folder in the include options?

As for the porting of TensorFlow for Microcontroller to C6748 DSP (Options 3), the porting process after install the library is just to include the folder in the include options?

+1 Jianzhong Xu over 4 years ago in reply to Zhi Sheng

TI__Guru 60890 points

For option 2, you'll need to have Sitara devices (AM335x, AM437x, AM57x, etc). Please follow Processor-SDK Linux TensorFlow Lite for detailed information.

For option 3, you'll need to compile the source code for C6748 which I haven't tried and TI has no support for.

Thanks,

Jianzhong

Processors

Processors forum

TMS320C6748: How to Port or Install Machine Learning?