Comment by kerasteam

That's right, if the model is backend-agnostic you can train it with a PyTorch training loop and then reload it and use it with TF ecosystem tools, like serve it with TF-Serving or export it to Coral TPU.