Inferencing 2.3 trained Model using the 2.5 release

[mohamadmansourX]

2.3 Release Trained Model - 2.5 Release Code results mismatch. After training my recognition model using the 2.3 release and testing it successfully, when trying to use the same weights on the new 2.5 release the results gets disrupted; having some comparable results on some cases while very bad ones on others. Was I supposed to do something with my weights before migrating to the latest release?