Can't load keras model with more than 1 metric

Question

I have issues with tensforflow when I try to load a model which contains more than 1 metric. It gives the following error:

ValueError: Unable to restore custom object of type _tf_keras_metric currently. Please make sure that the layer implements get_config and from_config when saving. In addition, please use the custom_objects arg when calling load_model() .

I have tried to search for solutions, but I can only find examples with 1 metric, which also works for me, but I need mulitple metrics. Hope you guys can help!

My code:

METRICS = [
 
      keras.metrics.BinaryAccuracy(threshold = 0.5),
      tfa.metrics.HammingLoss(mode='multilabel'),
      tfa.metrics.F1Score(num_classes=4, threshold=0.5)      
          
]

VOCAB_SIZE = 25000
encoder = tf.keras.layers.experimental.preprocessing.TextVectorization(max_tokens=VOCAB_SIZE, output_mode='int', pad_to_max_tokens = True)

encoder.adapt(X_train)

model = tf.keras.Sequential([encoder,
        tf.keras.layers.Embedding(input_dim=len(encoder.get_vocabulary())+1, output_dim=64, mask_zero=True),
        tf.keras.layers.Dropout(0.5),                     
        tf.keras.layers.GlobalAveragePooling1D(),
        tf.keras.layers.Dropout(0.5),
        tf.keras.layers.Dense(32, activation='sigmoid', activity_regularizer=tf.keras.regularizers.L2(0.005)),
        tf.keras.layers.Dense(4, activation='sigmoid')      
                            
                             ])

model.compile(loss=tf.keras.losses.BinaryCrossentropy(from_logits=False),                    
              optimizer=tf.keras.optimizers.Adam(learning_rate = 0.0005),
              metrics= METRICS),
              
history = model.fit(X_train, label_train,
                    epochs=2,
                    validation_split = 0.2,
                    batch_size = 32,
                    verbose = 1,
                    shuffle = True)

# --- Save trained model --- #
model.save('CNN_model_fit_1.2', save_format = 'tf')

# --- Load model --- #
from keras.models import load_model
def BinaryAccuracy(label_test, test_predictions):
    return 1
def HammingLoss(label_test, test_predictions):
    return 1
def F1Score(label_test, test_predictions):
    return 1

model_new = load_model("CNN_model_fit_1.2", custom_objects={'binary_accuracy':BinaryAccuracy,'hamming_loss':HammingLoss,'f1_score':F1Score})

model_new.get_weights()
pred = model_new.predict(X_test)

Answer 1

I cannot test your code because you have not provided shape and size of X_train but here is an idea:

Tensorflow addons is a separate library from Tensorflow , hence its metrics can be thought of as custom objects.

#Add tf.metrics before the name of the metrics
model_new = load_model("CNN_model_fit_1.2", custom_objects={'hamming_loss': tfa.metrics.HammingLoss,'f1_score': tfa.metrics.F1Score})