3 ms·I’m curious, why would you want to define the loss in the model and not in the trainer?by p1esk 6y agoI’m curious, why would you want to define the loss in the model and not in the trainer?ipsum2 6y agoIt's standard TensorFlow practice. e.g. `model.compile(loss=loss)`p1esk 6y agoThe way things are done in TF can hardly ever be considered good practices.ipsum2 6y agoI don't disagree with you, which is why I use PyTorch instead of TensorFlow. But ZenML is very tied to the TF ecosystem.