• State TI Thinks Resolved
  • Locked Locked
  • Replies 1 reply
  • Answers 1 answer
  • Subscribers 97 subscribers
  • Views 231 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TDA4VM: Implementing TIDL using PSDK on EVM

Aditi Ramadwar
Prodigy 20 points
Part Number: TDA4VM

A. TIDL-RT

Following this particular instruction set:

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/08_02_00_05/exports/docs/tidl_j721e_08_02_00_11/ti_dl/docs/user_guide_html/md_tidl_user_model_deployment.html

The demo is carried out properly and the required output files are being generated after compilation.

1. How to use those output files to run the same model on the EVM?

2. If I want to use one of the other models present in "ti-processor-sdk-rtos-j721e-evm-08_02_00_05/tidl_j721e_08_02_00_11/ti_dl/test/testvecs/config/import/public/onnx/" where should I download the particular model from since the "../../test/testvecs/models/public/onnx/" folder is empty?

3. If I want to use a custom model then how should I generate a .txt file similar to that present in "ti-processor-sdk-rtos-j721e-evm-08_02_00_05/tidl_j721e_08_02_00_11/ti_dl/test/testvecs/config/import/public/onnx". Considering I already have the model and its artifacts generated.

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/08_02_00_05/exports/docs/tidl_j721e_08_02_00_11/ti_dl/docs/user_guide_html/md_tidl_model_import.html

Should the above link be referenced every time for a new custom model?

B. Edgeai TIDL Tools

The other method for generating inference: https://github.com/TexasInstruments/edgeai-tidl-tools is used

Since this git repository is outside of the PSDK, which files from the repository are to be moved to which folder in the PSDK after compilation of a model on PC?

  • 0
    TI__Genius 9065 points

    Hi

    1. Refer to the following section : Steps to run sample application on EVM : https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/08_02_00_05/exports/docs/tidl_j721e_08_02_00_11/ti_dl/docs/user_guide_html/md_tidl_sample_test.html

    2. You need to place your trained model at the specified location. Or download whichever model you need.

    3. You can copy any of the existing config files and modify based on your need. You need to understand the usage of parameter from the documentation link you are referring to.

    Most of the times, replicating an existing txt file, and modifying the paths to input data and model should work, but I would suggest going over the doc so you are aware of all the options available.

    Edgeai_tidl_tools is a self-sufficient repo and you need not copy anything from PSDK. The required tools from PSDK are installed/downloaded as part of the repo setup. You can run both compilation (on PC) and inference (on PC/EVM) using this repo, so no need to copy anything back to PSDK. You can just use this repo for inference as well. Please refer to the doc in this repo.

    Regards,

    Anand