multilabel does not work for audio event detection

[louisfh]

It would be nice to have a multilabel option for the audio event detection models.

Setting
multi_label: True

In the config.yaml for a YAMNet Audio Event Detection model gives the following stack trace:

Error executing job with overrides: []
Traceback (most recent call last):
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/ENV/lib/python3.10/site-packages/clearml/binding/hydra_bind.py", line 230, in _patched_task_function
    return task_function(a_config, *a_args, **a_kwargs)
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/stm32ai_main.py", line 594, in main
    process_mode(mode=mode,
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/stm32ai_main.py", line 493, in process_mode
    chain_tbqeb(cfg=configs, train_ds=train_ds, valid_ds=valid_ds,
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/stm32ai_main.py", line 303, in chain_tbqeb
    trained_model_path = train(cfg=cfg, train_ds=train_ds, valid_ds=valid_ds,
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/src/training/train.py", line 208, in train
    model, _ = _load_model_to_train(cfg, model_path=cfg.general.model_path, num_classes=num_classes)
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/src/training/train.py", line 50, in _load_model_to_train
    model = get_model(
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/src/utils/models_mgt.py", line 88, in get_model
    model = yamnet.get_model(n_classes=num_classes,
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/src/models/yamnet/yamnet.py", line 143, in get_model
    yamnet = _get_transfer_learning_model(n_classes=n_classes,
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/src/models/yamnet/yamnet.py", line 91, in _get_transfer_learning_model
    yamnet = add_head(backbone=permuted_backbone,
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/stm32ai-modelzoo-services/audio_event_detection/src/models/model_utils.py", line 59, in add_head
    out = tf.keras.layers.Dense(units=n_classes, activation=activation,
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/ENV/lib/python3.10/site-packages/keras/layers/core/dense.py", line 118, in __init__
    self.activation = activations.get(activation)
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/ENV/lib/python3.10/site-packages/tensorflow/python/util/traceback_utils.py", line 153, in error_handler
    raise e.with_traceback(filtered_tb) from None
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/ENV/lib/python3.10/site-packages/keras/activations.py", line 595, in get
    return deserialize(identifier)
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/ENV/lib/python3.10/site-packages/keras/activations.py", line 555, in deserialize
    return deserialize_keras_object(
  File "/media/auk/projects/lfh/FROG6_LoRa_frogs/ENV/lib/python3.10/site-packages/keras/utils/generic_utils.py", line 709, in deserialize_keras_object
    raise ValueError(
ValueError: Unknown activation function: sigmoïd. Please ensure this object is passed to the `custom_objects` argument. See https://www.tensorflow.org/guide/keras/save_and_serialize#registering_the_custom_object for details.