使用度量“tf.keras.metrics.AUC”进行训练后，如何在 Keras 中使用“load_model”？

Question

After training using that metric on my model.fit() . When I try to load the model using "load_model". It don't recognize the metric AUC, so I add it on custom_objects={"auc":AUC} .

I get this error:

ValueError: Unknown metric function: {'class_name': 'AUC'... returning all the thresholds used and more information about the metric.

The code:

model.compile(..., metrics=["accuracy", AUC(name="auc", curve="PR")]
load_model(checkpoint, custom_objects={"auc":AUC(name="auc")})

Answer 1

I already answered. When you have no defined metrics, you can load the model without compile it. But adding the custom_objects like this:

he code:

model.compile(..., metrics=["accuracy", AUC(name="auc", curve="PR")] load_model(checkpoint, custom_objects={"auc":AUC(name="auc")}, compile=False)