How to set up DWN param.yaml

Emanuil Antonov

Part Number: TDA4VM

Hello, experts

Is there tool or manual how to generate param.yaml for custom model. I'm using https://github.com/TexasInstruments/edgeai-benchmark/blob/master/examples/configs/yaml/param_classification.yaml as reference.

over 2 years ago

0 Pratik Kedar over 2 years ago

TI__Mastermind 24041 points

Hi,

could you elaborate more on this issue ?

Are you trying to create param.yaml file for custom compiled model artifacts ?

Regards,

Pratik

0 Emanuil Antonov over 2 years ago in reply to Pratik Kedar

Prodigy 110 points

Yes i have artifacts already done with edgeai-tidl-tools but i don't have any yaml files created.
Regards,
Emanuil

0 Pratik Kedar over 2 years ago in reply to Emanuil Antonov

TI__Mastermind 24041 points

Hi,

We will get back to you on this issue by Monday.

Regards,

Pratik

0 Pratik Kedar over 2 years ago in reply to Pratik Kedar

TI__Mastermind 24041 points

Could you please share directory structure for custom compiled model artifact directory ?

You didn't see param.yaml file as part of generated artifacts ? is my understanding correct ?

Regards,

Pratik

0 Emanuil Antonov over 2 years ago in reply to Pratik Kedar

Prodigy 110 points

Yes, i dont't see any .yaml files. Output i got is:
├── intmodel
│   ├── bufinfolog.csv
│   ├── bufinfolog.txt
│   └── perfSimInfo.bin
├── intmodel.bin
├── intmodel.bin.layer_info.txt
├── intmodel.bin_netLog.txt
├── intmodel.bin_paramDebug.csv
├── intmodel.bin.svg
├── intmodel_io_1.bin
├── model_supported_layer_int.txt.perf_sim_config.txt
└── model_supported_layer_int.txt.qunat_stats_config.txt
command i used was:
TDA4_LINUX_RTOS/tidl_j721e_08_06_00_10/ti_dl/test/testvecs/config/import/public/tflite/model_supported_layer_int.txt

0 Pratik Kedar over 2 years ago in reply to Emanuil Antonov

TI__Mastermind 24041 points

Hi,

The param.yaml file is generated as outcome of model compilation process.

Could you re generate the model artifacts ?

Also please share the detailed process which you have followed to generate model artifacts along with TI documentation links.

Regards,

Pratik

0 Emanuil Antonov over 2 years ago in reply to Pratik Kedar

Prodigy 110 points

Hi,
> Could you re generate the model artifacts ?
Yes, the output is the same.

> Also please share the detailed process which you have followed to generate model artifacts along with TI documentation links
I follow documentation:
https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/08_00_00_12/exports/docs/tidl_j7_08_00_00_10/ti_dl/docs/user_guide_html/md_tidl_model_import.html
I write a <Configuration file> ---> /model_supported_layer_int.txt, according to the model which i wanna to import.

Regards,

Emanuil

0 Pratik Kedar over 2 years ago in reply to Emanuil Antonov

TI__Mastermind 24041 points

Hi,

Can you state your example use case ?

Are you trying to build model artifacts that can be consumed on target hardware for inferencing part ?

Regards,

Pratik

Processors

Processors forum

How to set up DWN param.yaml