demos/musicgen_colab.py

import argparse
import logging
import os
from pathlib import Path
import subprocess as sp
import sys
import time
import typing as tp
from tempfile import NamedTemporaryFile, gettempdir
from einops import rearrange
import torch
import gradio as gr
from audiocraft.data.audio_utils import convert_audio
from audiocraft.data.audio import audio_write
from audiocraft.models.encodec import InterleaveStereoCompressionModel
from audiocraft.models import MusicGen, MultiBandDiffusion
import multiprocessing as mp
import warnings

os.putenv("HF_HUB_ENABLE_HF_TRANSFER","1")
os.environ["SAFETENSORS_FAST_GPU"] = "1"

torch.backends.cuda.matmul.allow_tf32 = False
torch.backends.cuda.matmul.allow_bf16_reduced_precision_reduction = False
torch.backends.cuda.matmul.allow_fp16_reduced_precision_reduction = False
torch.backends.cudnn.allow_tf32 = False
torch.backends.cudnn.deterministic = False
torch.backends.cudnn.benchmark = False
# torch.backends.cuda.preferred_blas_library="cublas"
# torch.backends.cuda.preferred_linalg_library="cusolver"
torch.set_float32_matmul_precision("highest")

class FileCleaner:
    def __init__(self, file_lifetime: float = 3600):
        self.file_lifetime = file_lifetime
        self.files = []
    def add(self, path: tp.Union[str, Path]):
        self._cleanup()
        self.files.append((time.time(), Path(path)))
    def _cleanup(self):
        now = time.time()
        for time_added, path in list(self.files):
            if now - time_added > self.file_lifetime:
                if path.exists():
                    path.unlink()
                self.files.pop(0)
            else:
                break
                
file_cleaner = FileCleaner()

def convert_wav_to_mp4(wav_path, output_path=None):
    """Converts a WAV file to a waveform MP4 video using ffmpeg."""
    if output_path is None:
        # Create output path in the same directory as the input
        output_path = Path(wav_path).with_suffix(".mp4")
    try:
        command = [
            "ffmpeg",
            "-y",  # Overwrite output file if it exists
            "-i", str(wav_path),
            "-filter_complex",
            "[0:a]showwaves=s=1280x202:mode=line,format=yuv420p[v]",  # Waveform filter
            "-map", "[v]",
            "-map", "0:a",
            "-c:v", "libx264",  # Video codec
            "-c:a", "aac",       # Audio codec
            "-preset", "fast", # Important, don't do veryslow.
            str(output_path),
        ]
        process = sp.run(command, capture_output=True, text=True, check=True)
        return str(output_path)
    except sp.CalledProcessError as e:
        print(f"Error in ffmpeg conversion: {e}")
        print(f"ffmpeg stdout: {e.stdout}")
        print(f"ffmpeg stderr: {e.stderr}")
        raise  # Re-raise the exception to be caught by Gradio

def model_worker(model_name: str, task_queue: mp.Queue, result_queue: mp.Queue):
    """
    Persistent worker process (used when NOT running as a daemon).
    """
    try:
        device = 'cuda' if torch.cuda.is_available() else 'cpu'
        model = MusicGen.get_pretrained(model_name, device=device)
        mbd = MultiBandDiffusion.get_mbd_musicgen(device=device)
        while True:
            task = task_queue.get()
            if task is None:
                break
            task_id, text, melody, duration, use_diffusion, gen_params = task
            try:
                model.set_generation_params(duration=duration, **gen_params)
                target_sr = model.sample_rate
                target_ac = 1
                processed_melody = None
                if melody:
                    sr, melody_data = melody
                    melody_tensor = torch.from_numpy(melody_data).to(device).float().t()
                    if melody_tensor.ndim == 1:
                        melody_tensor = melody_tensor.unsqueeze(0)
                    melody_tensor = melody_tensor[..., :int(sr * duration)]
                    processed_melody = convert_audio(melody_tensor, sr, target_sr, target_ac)
                if processed_melody is not None:
                    output, tokens = model.generate_with_chroma(
                        descriptions=[text],
                        melody_wavs=[processed_melody],
                        melody_sample_rate=target_sr,
                        progress=True,
                        return_tokens=True
                    )
                else:
                    output, tokens = model.generate([text], progress=True, return_tokens=True)
                output = output.detach().cpu()
                if use_diffusion:
                    if isinstance(model.compression_model, InterleaveStereoCompressionModel):
                        left, right = model.compression_model.get_left_right_codes(tokens)
                        tokens = torch.cat([left, right])
                    outputs_diffusion = mbd.tokens_to_wav(tokens)
                    if isinstance(model.compression_model, InterleaveStereoCompressionModel):
                        assert outputs_diffusion.shape[1] == 1  # output is mono
                        outputs_diffusion = rearrange(outputs_diffusion, '(s b) c t -> b (s c) t', s=2)
                    outputs_diffusion = outputs_diffusion.detach().cpu()
                    result_queue.put((task_id, (output, outputs_diffusion)))
                else:
                    result_queue.put((task_id, (output, None)))
            except Exception as e:
                result_queue.put((task_id, e))
    except Exception as e:
        result_queue.put((-1, e))

class Predictor:
    def __init__(self, model_name: str, depth: str):
        self.model_name = model_name
        self.is_daemon = mp.current_process().daemon
        if self.is_daemon:
            # Running in a daemonic process (e.g., on Spaces)
            self.device = 'cuda' if torch.cuda.is_available() else 'cpu'
            self.model = MusicGen.get_pretrained(self.model_name, device=self.device, depth=depth)
            self.mbd = MultiBandDiffusion.get_mbd_musicgen(device=self.device)  # Load MBD here too
            self.current_task_id = 0 # Initialize task ID
        else:
            # Running in a non-daemonic process (e.g., locally)
            self.task_queue = mp.Queue()
            self.result_queue = mp.Queue()
            self.process = mp.Process(
                target=model_worker, args=(self.model_name, self.task_queue, self.result_queue)
            )
            self.process.start()
            self.current_task_id = 0
            self._check_initialization()

    def _check_initialization(self):
        """Check if the worker process initialized successfully (only in non-daemon mode)."""
        if not self.is_daemon:
            time.sleep(2)
            try:
                task_id, result = self.result_queue.get(timeout=3)
                if isinstance(result, Exception):
                    if task_id == -1:
                        raise RuntimeError("Model loading failed in worker process.") from result
            except:
                pass

    def predict(self, text, melody, duration, use_diffusion, **gen_params):
        """Submits a prediction task."""
        if self.is_daemon:
             # Directly perform the prediction (single-process mode)
            self.current_task_id +=1
            task_id = self.current_task_id
            try:
                self.model.set_generation_params(duration=duration, **gen_params)
                target_sr = self.model.sample_rate
                target_ac = 1
                processed_melody = None
                if melody:
                    sr, melody_data = melody
                    melody_tensor = torch.from_numpy(melody_data).to(self.device).float().t()
                    if melody_tensor.ndim == 1:
                        melody_tensor = melody_tensor.unsqueeze(0)
                    melody_tensor = melody_tensor[..., :int(sr * duration)]
                    processed_melody = convert_audio(melody_tensor, sr, target_sr, target_ac)
                if processed_melody is not None:
                    output, tokens = self.model.generate_with_chroma(
                        descriptions=[text],
                        melody_wavs=[processed_melody],
                        melody_sample_rate=target_sr,
                        progress=True,
                        return_tokens=True
                    )
                else:
                    output, tokens = self.model.generate([text], progress=True, return_tokens=True)
                output = output.detach().cpu()
                if use_diffusion:
                    if isinstance(self.model.compression_model, InterleaveStereoCompressionModel):
                        left, right = self.model.compression_model.get_left_right_codes(tokens)
                        tokens = torch.cat([left, right])
                    outputs_diffusion = self.mbd.tokens_to_wav(tokens)
                    if isinstance(self.model.compression_model, InterleaveStereoCompressionModel):
                        assert outputs_diffusion.shape[1] == 1  # output is mono
                        outputs_diffusion = rearrange(outputs_diffusion, '(s b) c t -> b (s c) t', s=2)
                    outputs_diffusion = outputs_diffusion.detach().cpu()
                    return task_id, (output, outputs_diffusion) #Return the task id.
                else:
                  return task_id, (output, None)
            except Exception as e:
                return task_id, e
        else:
            # Use the multiprocessing queue (multi-process mode)
            self.current_task_id += 1
            task = (self.current_task_id, text, melody, duration, use_diffusion, gen_params)
            self.task_queue.put(task)
            return self.current_task_id, (None, None)

    def get_result(self, task_id):
        """Retrieves the result of a prediction task."""
        if self.is_daemon:
            # Results are returned directly by 'predict' in daemon mode
            result_id, result = task_id, task_id #predictor return (task_id, results)
        else:
            # Get result from the queue (multi-process mode)
            while True:
                result_task_id, result = self.result_queue.get()
                if result_task_id == task_id:
                    break  # Found the correct result
        if isinstance(result, Exception):
            raise result
        return result

    def shutdown(self):
        """Shuts down the worker process (if running)."""
        if not self.is_daemon and self.process.is_alive():
            self.task_queue.put(None)
            self.process.join()

_default_model_name = "facebook/musicgen-melody"

def predict_full(model, model_path, depth, use_mbd, text, melody, duration, topk, topp, temperature, cfg_coef):
    # Initialize Predictor *INSIDE* the function
    predictor = Predictor(model, depth)
    task_id, (wav, diffusion_wav) = predictor.predict( # Unpack directly!
        text=text,
        melody=melody,
        duration=duration,
        use_diffusion=use_mbd,
        top_k=topk,
        top_p=topp,
        temperature=temperature,
        cfg_coef=cfg_coef,
    )
    # 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], 44100, strategy="loudness", #hardcoded sample rate
            loudness_headroom_db=16, loudness_compressor=True, add_suffix=False
        )
        wav_paths.append(file.name)
        # Make and clean up video:
        video_path = convert_wav_to_mp4(file.name)
        video_paths.append(video_path)
        file_cleaner.add(file.name)
        file_cleaner.add(video_path)
    # 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], 44100, strategy="loudness", #hardcoded sample rate
                loudness_headroom_db=16, loudness_compressor=True, add_suffix=False
            )
            wav_paths.append(file.name)
            # Make and clean up video:
            video_path = convert_wav_to_mp4(file.name)
            video_paths.append(video_path)
            file_cleaner.add(file.name)
            file_cleaner.add(video_path)
    # Shutdown predictor to prevent hanging processes!
    if not predictor.is_daemon: # Important!
        predictor.shutdown()
    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

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
                    depth = gr.Radio(["float32", "bfloat16", "float16"],
                                     label="Model Precision", value="float32", interactive=True)
                    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, depth, 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.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. facebook/musicgen-large -- a 3.3B transformer decoder conditioned on text only.
            5. facebook/musicgen-melody-large -- a 3.3B transformer decoder conditioned on text and melody.
            6. facebook/musicgen-stereo-small -- a 300M transformer decoder conditioned on text only, fine tuned for stereo output.
            7. facebook/musicgen-stereo-medium -- a 1.5B transformer decoder conditioned on text only, fine tuned for stereo output.
            8. facebook/musicgen-stereo-melody -- a 1.5B transformer decoder conditioned on text and melody, fine tuned for stereo output.
            9. facebook/musicgen-stereo-large -- a 3.3B transformer decoder conditioned on text only, fine tuned for stereo output.
           10. facebook/musicgen-stereo-melody-large -- a 3.3B transformer decoder conditioned on text and melody, fine tuned for stereo output.

            We also present two way of decoding the audio tokens:
            1. Use the default GAN based compression model.  It can suffer from artifacts especially
                for crashes, snares etc.
            2. Use [MultiBand Diffusion](https://arxiv.org/abs/2308.02560).  Should improve the audio quality,
                at an extra computational cost.  When this is selected, we provide both the GAN based decoded
                audio, and the one obtained with MBD.

            See [github.com/facebookresearch/audiocraft](https://github.com/facebookresearch/audiocraft/blob/main/docs/MUSICGEN.md)
            for more details.
            """
        )

        interface.queue().launch(**launch_kwargs)

if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument(
        '--listen',
        type=str,
        default='0.0.0.0' if 'SPACE_ID' in os.environ else '127.0.0.1',
        help='IP to listen on for connections to Gradio',
    )
    parser.add_argument(
        '--username', type=str, default='', help='Username for authentication'
    )
    parser.add_argument(
        '--password', type=str, default='', help='Password for authentication'
    )
    parser.add_argument(
        '--server_port',
        type=int,
        default=0,
        help='Port to run the server listener on',
    )
    parser.add_argument(
        '--inbrowser', action='store_true', help='Open in browser'
    )
    parser.add_argument(
        '--share', action='store_true', help='Share the gradio UI'
    )
    args = parser.parse_args()
    launch_kwargs = {}
    launch_kwargs['server_name'] = args.listen
    if args.username and args.password:
        launch_kwargs['auth'] = (args.username, args.password)
    if args.server_port:
        launch_kwargs['server_port'] = args.server_port
    if args.inbrowser:
        launch_kwargs['inbrowser'] = args.inbrowser
    launch_kwargs['share'] = True
    logging.basicConfig(level=logging.INFO, stream=sys.stderr)
    # Added predictor shutdown
    try:
        ui_full(launch_kwargs)
    finally:
        if _predictor is not None:
            _predictor.shutdown()