How to get all scores from the Yamnet tutorial, instead of just the top score?

Hi, I’m new to Tensorflow and I’m explorign this tutorial :

Since I’m doing a little project that is about figuring out what kind of sounds an audio clip contains. Since my audio dataset is not labelled, and I’m only interested in to knowing the top or more broad classes, the question is:

On my demo audio clip, there are two elements on it : a person talking and a dog barking at different points in time. When I run the example :

scores, embeddings, spectrogram = yamnet_model(testing_wav_data)

class_scores = tf.reduce_mean(scores, axis=0)

top_class = tf.argmax(class_scores)

inferred_class = class_names[top_class]

print(f'The main sound is: {inferred_class}')

print(f'The embeddings shape: {embeddings.shape}')
The main sound is: Speech
The embeddings shape: (109, 1024)

This is correctly predicted, since there is actual speech for most of the time of the clip, maybe 65% of the time. But then at the end there is a dog barking. My question is, are these scores for the dog barking also present in the prediction results ? How can I map all the results predicted and not just the top_class So I can check if there is a high score, although in the second position, for “animal” ?