Implementation Differences between SGD Tensorflow and PyTorch SGD

Question

Some issues I am getting while training with identical code ported from Pytorch to TF2.

• Model.fit converges in a completely different manner than Gradient Tape. (An