You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
A Keras usage pattern that allows for maximum flexibility when defining arbitrary losses and metrics (that don't match the usual signature) is the "endpoint layer" pattern. It works like this: https://colab.research.google.com/drive/1zzLcJ2A2qofIvv94YJ3axRknlA6cBSIw …pic.twitter.com/v3YtMrrcQb
In short, you use `add_loss`/`add_metric` inside an "endpoint layer" that also has access to model targets. The layer then returns the inference-time predictions. You compile without an external "loss" argument, and you fit with a dictionary of data that contains the targets.
Of course logistic regression is a basic case that doesn't actually need this advanced pattern. But endpoint layers will work every time, even when you have losses & metrics that don't match the usual `fn(y_true, y_pred, sampl_weight)` signature that is required in `compile`.
A very good method to customize our neural networks. This method can be generalized for other machine learning techniques too.
How does this work during inference, when targets are not available? Do we only add loss and metric when the targets are not None.
You need to redefine your model for inference. This is pretty easy since you can reuse almost all the logic from the training model.
cc @notwaldorf for all your custom crazy losses
I assume this is the new preferred way of using complex losses over using the Function layer, as in the OCR example?
Nice. Actually we use sometime like this for CTC loss (targets as inputs, add_loss()).
Is this feature available in TensorFlow Eager model without compiling?
Tried this, getting an error at self.add_metric saying that the attribute doesn't exist...
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
