Backbone model weight parameters have been changing each running..

[omeralierdemir]

Hi, I try to convert resnet feature extractor backbone to tesnorrt.engine file. I have written Resnet50 backbone with tensorrt library. Then I have compared resnet50 backbone parameters (your python backbone) with my tensorrt feature extractor model parameters to check the accuracy of the tensorrt model result. Then I noticed that the weights of your feature axractor resnet50 model parameters changes with every run. I have checked your dimp50.pth file load code. But I have not saw any mistake. Do you have any idea this situation? Why changes your feature exractor backbone resnet50 model parameters every running step ?

[goutamgmb]

Hi,

That is strange. Could you post the code you are using to check the model weights?