TDA4VM: Is it possible to inference model using edge AI and the bin file imported by TIDL importer?

Liu Tsing-Yue

Intellectual 291 points

Part Number: TDA4VM

Hi, i'm using edge-AI tools to inference compiled onnx models.

The quantized model (.bin file) was imported by TIDL importer

However, the model structures are different, some data_convert layers are in the model compiled by edge-AI.

Is it possible to inference model using edge AI and the bin file imported by TIDL importer? Thanks!

over 1 year ago

0 Liu Tsing-Yue over 1 year ago

Intellectual 291 points

model_visualization.zip

Here's a comparison of visualized models (compiled by edge-AI vs imported by TIDL importer)

+1 Pratik Kedar over 1 year ago in reply to Liu Tsing-Yue

TI__Mastermind 24041 points

Hi,

As I understand this query, you have compiled the onnx model for TDA4VM SoC, and you want to perform model inference using the same ?

On edgeai-linux-sdk you can refer to this FAQ : https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1228053/faq-sk-tda4vm-how-to-do-inferencing-with-custom-compiled-model-on-target

Please note that on edgeai-linux sdk flow mentioned above needs param.yaml file to be present as part of artifacts (This can be achieved by compiling your model using edgeai benchmark repos here : https://github.com/TexasInstruments/edgeai-benchmark)

If you have generated model artifacts using edgeai tidl tools repo : https://github.com/TexasInstruments/edgeai-tidl-tools

You can refer the doc here : https://github.com/TexasInstruments/edgeai-tidl-tools#benchmark-on-ti-soc

Processors

Processors forum

TDA4VM: Is it possible to inference model using edge AI and the bin file imported by TIDL importer?