app.py

import gradio as gr
import librosa
from transformers import pipeline

pipe = pipeline("audio-classification", model="lewtun/distilhubert-finetuned-gtzan-og")


def classify_audio(filepath):
    audio, sampling_rate = librosa.load(filepath, sr=16_000)
    preds = pipe(audio)
    outputs = {}
    for p in preds:
        outputs[p["label"]] = p["score"]
    return outputs


label = gr.outputs.Label()

demo = gr.Interface(fn=classify_audio, inputs=gr.Audio(type="filepath"), outputs=label)
demo.launch()