server.py

import warnings
import argparse
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from typing import List, Optional
import torch
from torch.cuda import memory_allocated, memory_reserved
from audiocraft.models import musicgen
import numpy as np
import io
from fastapi.responses import StreamingResponse, JSONResponse
from scipy.io.wavfile import write as wav_write
import uvicorn
import psutil

warnings.simplefilter('ignore')

# Parse command line arguments
parser = argparse.ArgumentParser(description="Music Generation Server")
parser.add_argument("--model", type=str, default="musicgen-stereo-small", help="Pretrained model name")
parser.add_argument("--device", type=str, default="cuda", help="Device to load the model on")
parser.add_argument("--duration", type=int, default=10, help="Duration of generated music in seconds")
parser.add_argument("--host", type=str, default="0.0.0.0", help="Host to run the server on")
parser.add_argument("--port", type=int, default=8000, help="Port to run the server on")

args = parser.parse_args()

# Initialize the FastAPI app
app = FastAPI()

# Build the model name based on the provided arguments
if args.model.startswith('facebook/'):
    args.model_name = args.model
else:
    args.model_name = f"facebook/{args.model}"

# Load the model with the provided arguments
try:
    musicgen_model = musicgen.MusicGen.get_pretrained(args.model_name, device=args.device)
    model_loaded = True
except Exception as e:
    musicgen_model = None
    model_loaded = False

class MusicRequest(BaseModel):
    prompts: List[str]
    duration: Optional[int] = 10  # Default duration is 10 seconds if not provided

@app.post("/generate_music")
def generate_music(request: MusicRequest):
    if not model_loaded:
        raise HTTPException(status_code=500, detail="Model is not loaded.")
    
    try:
        musicgen_model.set_generation_params(duration=request.duration)
        result = musicgen_model.generate(request.prompts, progress=False)
        result = result.squeeze().cpu().numpy().T
        
        sample_rate = musicgen_model.sample_rate
        
        buffer = io.BytesIO()
        wav_write(buffer, sample_rate, result)

        buffer.seek(0)
        
        return StreamingResponse(buffer, media_type="audio/wav")
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

@app.get("/health")
def health_check():
    cpu_usage = psutil.cpu_percent(interval=1)
    ram_usage = psutil.virtual_memory().percent
    stats = {
        "server_running": True,
        "model_loaded": model_loaded,
        "cpu_usage_percent": cpu_usage,
        "ram_usage_percent": ram_usage
    }
    if args.device == "cuda" and torch.cuda.is_available():
        gpu_memory_allocated = memory_allocated()
        gpu_memory_reserved = memory_reserved()
        stats.update({
            "gpu_memory_allocated": gpu_memory_allocated,
            "gpu_memory_reserved": gpu_memory_reserved
        })
    return JSONResponse(content=stats)

if __name__ == "__main__":
    uvicorn.run(app, host=args.host, port=args.port)