from transformers import pipeline import torch from transformers import pipeline from transformers.pipelines.audio_utils import ffmpeg_read import gradio as gr import gradio as gr device = 0 if torch.cuda.is_available() else "cpu" MODEL_ID = "jvalero/wav2vec2-base-music_genre_classifier" pipe = pipeline( task="audio-classification", model=MODEL_ID, chunk_length_s=30, device=device, ) id2label = { "LABEL_0" : 'drumbass', "LABEL_1" : 'dubtechno', "LABEL_2": 'dupstep', "LABEL_3":'hardcore_breaks', "LABEL_4": 'house', "LABEL_5":'psytrance', "LABEL_6" : 'techno', "LABEL_7":'ukgarage' } def get_edm(filepath): output = pipe( filepath, max_new_tokens=256, chunk_length_s=30, batch_size=8, ) return id2label[output[0]["label"]] demo = gr.Blocks() demo = gr.Blocks() file_transcribe = gr.Interface( fn=get_edm, inputs=[ gr.Audio(sources="upload", label="Audio file", type="filepath"), ], outputs="label", title="EDM genre classificator", description=( "Get the genre of your song! Demo uses the" f" checkpoint [{MODEL_ID}](https://huggingface.co/{MODEL_ID}) and 🤗 Transformers to get the condition of audio files" " of arbitrary length. \nThe audio will be classified into one of the following: ['drumbass', 'dubtechno', 'dupstep', 'hardcore_breaks', 'house', 'psytrance', 'techno', 'ukgarage']" ), examples=[ ["./example1.mp3"], ["./example2.mp3"], ["./example3.mp3"], ["./example4.mp3"], ["./example5.mp3"], ["./example6.mp3"], ], cache_examples=True, allow_flagging="never", ) with demo: gr.TabbedInterface([file_transcribe], ["Get Viny Condition"]) demo.launch()