Hi I'm working on a similar pipeline to convert my pytorch model to tensorflow lite (my aim is to use my NN on Google Coral TPU, which supports only TF-Lite). My pytorch network is a resnet18. I was wondering about the modified version of ONN-TF you mentioned in the README file. Did you modified the library by yourself? Or did you just download it somewhere as it is. Moreover, what's the reason of this customization? And last but not least, if it's possible, could you share that version?
I will probably move to use your converter. If i want to use it, do I need the modified version of ONN-TF, or could i just grab the last one from pip? Same question for the others libraries.
Thanks a lot for the availability of sharing your solution
have a nice day,
Francesco Montedori
Hi I'm working on a similar pipeline to convert my pytorch model to tensorflow lite (my aim is to use my NN on Google Coral TPU, which supports only TF-Lite). My pytorch network is a resnet18. I was wondering about the modified version of ONN-TF you mentioned in the README file. Did you modified the library by yourself? Or did you just download it somewhere as it is. Moreover, what's the reason of this customization? And last but not least, if it's possible, could you share that version?
I will probably move to use your converter. If i want to use it, do I need the modified version of ONN-TF, or could i just grab the last one from pip? Same question for the others libraries.
Thanks a lot for the availability of sharing your solution
have a nice day,
Francesco Montedori