Update demos/musicgen_app.py

demos/musicgen_app.py

@@ -154,4 +154,211 @@ class Predictor:
             result_task_id, result = self.result_queue.get()
             if result_task_id == task_id:
                 if isinstance(result, Exception):
-                    raise result
+                    raise result  # Re-raise the exception in the main process
+                return result # (wav, diffusion_wav) or (wav, None)
+
+    def shutdown(self):
+        """
+        Shuts down the worker process.
+        """
+        self.task_queue.put(None)  # Send sentinel value to stop the worker
+        self.process.join()  # Wait for the process to terminate
+
+
+# Global predictor instance
+_predictor = None
+
+def get_predictor(model_name:str = 'facebook/musicgen-melody'):
+    global _predictor
+    if _predictor is None:
+        _predictor = Predictor(model_name)
+    return _predictor
+
+def predict_full(model, model_path, use_mbd, text, melody, duration, topk, topp, temperature, cfg_coef):
+
+    predictor = get_predictor(model)
+    task_id = predictor.predict(
+        text=text,
+        melody=melody,
+        duration=duration,
+        use_diffusion=use_mbd,
+        top_k=topk,
+        top_p=topp,
+        temperature=temperature,
+        cfg_coef=cfg_coef,
+    )
+
+    wav, diffusion_wav = predictor.get_result(task_id)
+
+    # Save and return audio files
+    wav_paths = []
+    video_paths = []
+
+    # Save standard output
+    with NamedTemporaryFile("wb", suffix=".wav", delete=False) as file:
+        audio_write(
+            file.name, wav[0], 32000, strategy="loudness", #hardcoded sample rate
+            loudness_headroom_db=16, loudness_compressor=True, add_suffix=False
+        )
+        wav_paths.append(file.name)
+        video_paths.append(make_waveform(file.name)) # Make and clean up video
+        file_cleaner.add(file.name)
+
+    # Save MBD output if used
+    if diffusion_wav is not None:
+        with NamedTemporaryFile("wb", suffix=".wav", delete=False) as file:
+            audio_write(
+                file.name, diffusion_wav[0], 32000, strategy="loudness", #hardcoded sample rate
+                loudness_headroom_db=16, loudness_compressor=True, add_suffix=False
+            )
+            wav_paths.append(file.name)
+            video_paths.append(make_waveform(file.name)) # Make and clean up video
+            file_cleaner.add(file.name)
+
+    if use_mbd:
+         return video_paths[0], wav_paths[0], video_paths[1], wav_paths[1]
+    return video_paths[0], wav_paths[0], None, None
+
+
+def toggle_audio_src(choice):
+    if choice == "mic":
+        return gr.update(sources="microphone", value=None, label="Microphone")
+    else:
+        return gr.update(sources="upload", value=None, label="File")
+
+
+def toggle_diffusion(choice):
+    if choice == "MultiBand_Diffusion":
+        return [gr.update(visible=True)] * 2
+    else:
+        return [gr.update(visible=False)] * 2
+# --- Gradio UI ---
+
+def ui_full(launch_kwargs):
+    with gr.Blocks() as interface:
+        gr.Markdown(
+            """
+            # MusicGen
+            This is your private demo for [MusicGen](https://github.com/facebookresearch/audiocraft),
+            a simple and controllable model for music generation
+            presented at: ["Simple and Controllable Music Generation"](https://huggingface.co/papers/2306.05284)
+            """
+        )
+        with gr.Row():
+            with gr.Column():
+                with gr.Row():
+                    text = gr.Text(label="Input Text", interactive=True)
+                    with gr.Column():
+                        radio = gr.Radio(["file", "mic"], value="file",
+                                         label="Condition on a melody (optional) File or Mic")
+                        melody = gr.Audio(sources="upload", type="numpy", label="File",
+                                          interactive=True, elem_id="melody-input")
+                with gr.Row():
+                    submit = gr.Button("Submit")
+                    # _ = gr.Button("Interrupt").click(fn=interrupt, queue=False)  # Interrupt is now handled implicitly
+                with gr.Row():
+                    model = gr.Radio(["facebook/musicgen-melody", "facebook/musicgen-medium", "facebook/musicgen-small",
+                                      "facebook/musicgen-large", "facebook/musicgen-melody-large",
+                                      "facebook/musicgen-stereo-small", "facebook/musicgen-stereo-medium",
+                                      "facebook/musicgen-stereo-melody", "facebook/musicgen-stereo-large",
+                                      "facebook/musicgen-stereo-melody-large"],
+                                     label="Model", value="facebook/musicgen-melody", interactive=True)
+                    model_path = gr.Text(label="Model Path (custom models)", interactive=False, visible=False)  # Keep, but hide
+                with gr.Row():
+                    decoder = gr.Radio(["Default", "MultiBand_Diffusion"],
+                                       label="Decoder", value="Default", interactive=True)
+                with gr.Row():
+                    duration = gr.Slider(minimum=1, maximum=120, value=10, label="Duration", interactive=True)
+                with gr.Row():
+                    topk = gr.Number(label="Top-k", value=250, interactive=True)
+                    topp = gr.Number(label="Top-p", value=0, interactive=True)
+                    temperature = gr.Number(label="Temperature", value=1.0, interactive=True)
+                    cfg_coef = gr.Number(label="Classifier Free Guidance", value=3.0, interactive=True)
+            with gr.Column():
+                output = gr.Video(label="Generated Music")
+                audio_output = gr.Audio(label="Generated Music (wav)", type='filepath')
+                diffusion_output = gr.Video(label="MultiBand Diffusion Decoder", visible=False)
+                audio_diffusion = gr.Audio(label="MultiBand Diffusion Decoder (wav)", type='filepath', visible=False)
+
+        submit.click(
+            toggle_diffusion, decoder, [diffusion_output, audio_diffusion], queue=False
+        ).then(
+            predict_full,
+            inputs=[model, model_path, decoder, text, melody, duration, topk, topp, temperature, cfg_coef],
+            outputs=[output, audio_output, diffusion_output, audio_diffusion]
+        )
+        radio.change(toggle_audio_src, radio, [melody], queue=False, show_progress=False)
+
+        gr.Examples(
+            fn=predict_full,
+            examples=[
+                [
+                    "An 80s driving pop song with heavy drums and synth pads in the background",
+                    "./assets/bach.mp3",
+                    "facebook/musicgen-melody",
+                    "Default"
+                ],
+                [
+                    "A cheerful country song with acoustic guitars",
+                    "./assets/bolero_ravel.mp3",
+                    "facebook/musicgen-melody",
+                    "Default"
+                ],
+                [
+                    "90s rock song with electric guitar and heavy drums",
+                    None,
+                    "facebook/musicgen-medium",
+                    "Default"
+                ],
+                [
+                    "a light and cheerly EDM track, with syncopated drums, aery pads, and strong emotions",
+                    "./assets/bach.mp3",
+                    "facebook/musicgen-melody",
+                    "Default"
+                ],
+                [
+                    "lofi slow bpm electro chill with organic samples",
+                    None,
+                    "facebook/musicgen-medium",
+                    "Default"
+                ],
+                [
+                    "Punk rock with loud drum and power guitar",
+                    None,
+                    "facebook/musicgen-medium",
+                    "MultiBand_Diffusion"
+                ],
+            ],
+            inputs=[text, melody, model, decoder],
+            outputs=[output]
+        )
+        gr.Markdown(
+            """
+            ### More details
+
+            The model will generate a short music extract based on the description you provided.
+            The model can generate up to 30 seconds of audio in one pass.
+
+            The model was trained with description from a stock music catalog, descriptions that will work best
+            should include some level of details on the instruments present, along with some intended use case
+            (e.g. adding "perfect for a commercial" can somehow help).
+
+            Using one of the `melody` model (e.g. `musicgen-melody-*`), you can optionally provide a reference audio
+            from which a broad melody will be extracted.
+            The model will then try to follow both the description and melody provided.
+            For best results, the melody should be 30 seconds long (I know, the samples we provide are not...)
+
+            It is now possible to extend the generation by feeding back the end of the previous chunk of audio.
+            This can take a long time, and the model might lose consistency. The model might also
+            decide at arbitrary positions that the song ends.
+
+            **WARNING:** Choosing long durations will take a long time to generate (2min might take ~10min).
+            An overlap of 12 seconds is kept with the previously generated chunk, and 18 "new" seconds
+            are generated each time.
+
+            We present 10 model variations:
+            1. facebook/musicgen-melody -- a music generation model capable of generating music condition
+                on text and melody inputs. **Note**, you can also use text only.
+            2. facebook/musicgen-small -- a 300M transformer decoder conditioned on text only.
+            3. facebook/musicgen-medium -- a 1.5B transformer decoder conditioned on text only.
+            4