derrkater

@mohantym I should have mentioned it, but I use select ops syntax already. My conversion code is the same as you've posted and the problem persists :/

@mohantym the permissions are set now :) I did not add the dependency to the build.gradle file as I'm trying to run prediction on my mac first. It's first time...

@mohantym I haven't really tried iOS nor Android inference yet :) I just want to run this tflite checkpoint in python on my Mac using the code I've originally provided....

@mohantym I don't actually have a tensorflow model .py file as it was converted from pytorch via onnx format. I can provide a pb file with variables - is that...

I don't have any idea how to go about debugging such model :/ Is it maybe possible to catch intermediate steps of tflite forward pass to check where exactly it...

@karimnosseir inference in `pb` format works just fine and the problems begin only with tflite.

@mohantym @karimnosseir @sachinprasadhs is it possible for someone to take over the bug?

@mohantym @karimnosseir @sachinprasadhs is there any plan to go back to this problem? I'm stuck without support with this problem

@haozha111 the model always operates on this datatype (float32) and shape ([1,3,416,416]). Onnx version works just fine and it breaks after being transferred to tflite.

@haozha111 as I have mentioned before, I’m using select ops syntax and my conversion code is identical as yours. > @mohantym I don't actually have a tensorflow model .py file...