TensorFlow 1.x models no longer working properly in SDK >= 7.2

[Tarkkett]

Our team is trying to migrate to the newest SDK version, because it is required for the competition. The thing is, that earlier we had a fully working, custom tensorflow 1 object detection model and as soon as we updated the version, the model didn't work anymore (not detecting anything at all). Did the pre-trained model change? The same code works properly in SDK versions <7.2.(7.2 not included). We didn't find any breaking changes in the documentations of these newer versions :( Has anyone had this issue? If so, please let me know how to fix it.