Numbers are off when calculating cumulative average

Question

I'm trying to calculate the cumulative average (element-wise) of a numpy array in python. Here is the code:

class Prediction:

    # irrelevant code omitted

        cumulative_average = np.ones((1,494))
        for i, clip in enumerate(clip_paths):
            if glob.glob(f'{clip}.png'):
                async_to_sync(self.channel_layer.send)(self.channel_name, {
                    'type': 'websocket.send',
                    'text': f'skipping {clip}'})
                continue
            y,sr = librosa.load(clip, duration=3)
            
            mels = librosa.feature.melspectrogram(y=y, sr=sr)
            fig = plt.Figure()
            canvas = FigureCanvas(fig)
            p = plt.imshow(librosa.power_to_db(mels,ref=np.max))
            plt.axis('off')
            p.axes.get_xaxis().set_visible(False)
            p.axes.get_yaxis().set_visible(False)

            filename = clip.split('/')[-1].split('.mp3')[0]
            
            spectrogram_path = f'/tmp/{filename}.png'
            plt.savefig(spectrogram_path, bbox_inches='tight', pad_inches=0)

            spectrogram = PIL.Image.open(spectrogram_path)
            img = np.array(spectrogram)[:, :, :3]
            cumulative_average = (self.query_model(img) + (i * cumulative_average)) / (i + 1)
            
            # self.query_model(img) returns an array of shape (1,494)

            async_to_sync(self.channel_layer.send)(self.channel_name, {
                'type': 'websocket.send',
                'text': f'converted and got predictions for spectrogram image {spectrogram_path}'})

I omitted the irrelevant bits of the code, but comment if you need more information. The forumla for the cumulative average is on wikipedia: https://en.wikipedia.org/wiki/Moving_average#Cumulative_average

The numbers I'm getting are off. All of the numbers in the array are around .93

Can anyone help?