• State TI Thinks Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 1 answer
  • Subscribers 101 subscribers
  • Views 243 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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: Edge AI : Pytorch importing and post processing code snippet in the inference pipeline

abhilash sk
Intellectual 530 points
Part Number: TDA4VM

HI Champs , 

In the edgeai onnx_ep.py file can we have a post processing code snippet which has pytorch dependency as a post processing logic  , so this is working well on the Host x86 but can we run the same the same on EVM board ? IF not what additional installation has to be done to run the pytorch code snippet 

Setting : 

Edgeai sdk tag  : 09_02_07_00

System : Linux ubuntu 22.04

Model : Custom model 

  • 0
    TI__Mastermind 18570 points

    Hi,

    I have assigned this to the TIDL expert. Please expect a response soon.

    Regards,

    Fabiana

  • 0 in reply to Fabiana Jaimes
    TI__Mastermind 24041 points

    Hi,

    abhilash sk said:
    code snippet which has pytorch dependency as a post processing logic 

    Can you explain by above term ? this question lacks some clarity to explain the exact ask, it would be great if you throw some light on this.

    Post processing is comply dependent on custom application, users are welcome to bring there own logic. Coming to lib/package support to enable it on target side, personally i have never tried this use caes, however am hoping you can install python dependency on targe via pip tool as

    pip3 install <package-name> 

  • 0 in reply to Pratik Kedar
    Intellectual 530 points

    the question relates to if am having post processing logic which has pytorch dependency then can in incroporate that in the onnx_ep.py file code ? will it add any issues when we run it on the EVM target 

  • 0 in reply to abhilash sk
    TI__Mastermind 24041 points

    Hi,

    Here is my understanding about your request, correct me in case.

    abhilash sk said:
    if am having post processing logic which has pytorch dependency

    seems like you have code snippet for post processing with pytorch specific lib dependency (am i right here ?)

    You ask is that when i run same python script that works on host side now on target front how shall we cater to pytorch dependency ?

    As i pointed out in my previous response here,

    Pratik Kedar said:

    Post processing is comply dependent on custom application, users are welcome to bring there own logic. Coming to lib/package support to enable it on target side, personally i have never tried this use caes, however am hoping you can install python dependency on targe via pip tool as

    pip3 install <package-name> 

    You can try out this approach i have suggested, however am setting expectation that this is something we have not exercised.

    We have used PIL modules like ImageDraw for post processing implementation, and this is been verified at our end,

    here is the reference for https://github.com/TexasInstruments/edgeai-tidl-tools/blob/master/examples/osrt_python/common_utils.py#L299