feat: Migrate the deployment to the Gradio SDK, integrate the `spaces.GPU` decorator, and remove the Dockerfile.

Dockerfile

@@ -1,27 +0,0 @@
-# Use Python 3.12 to satisfy the 'perception-models' requirement
-FROM python:3.12
-
-# Set the working directory
-WORKDIR /code
-
-# Install system dependencies (ffmpeg is required for audio)
-RUN apt-get update && apt-get install -y ffmpeg && rm -rf /var/lib/apt/lists/*
-
-# Copy requirements and install Python dependencies
-COPY requirements.txt .
-RUN pip install --no-cache-dir --upgrade pip
-RUN pip install --no-cache-dir -r requirements.txt
-
-# Set up a user (Required by HF Spaces security)
-RUN useradd -m -u 1000 user
-USER user
-ENV HOME=/home/user \
-    PATH=/home/user/.local/bin:$PATH
-
-WORKDIR $HOME/app
-
-# Copy application files
-COPY --chown=user . $HOME/app
-
-# Start the app
-CMD ["python", "app.py"]

README.md

@@ -3,10 +3,13 @@ title: Sam Audio Webui
 emoji: 🎵
 colorFrom: indigo
 colorTo: pink
-sdk: docker
-app_port: 7860
+sdk: gradio
+sdk_version: 6.2.0
+app_file: app.py
 pinned: false
 license: apache-2.0
+fullWidth: true
+python_version: 3.11
 ---
 
 # SAM Audio WebUI
@@ -17,43 +20,11 @@ This Space hosts a WebUI for the **SAM Audio** model by Meta (Facebook), designe
 
 - **Model**: Uses `facebook/sam-audio-small` for a balance of performance and resource usage.
 - **ZeroGPU Support**: Optimized to run on Hugging Face ZeroGPU (A100/A10G) with automatic GPU handling.
-- **Dynamic Fallback**: 
-    - Attempts to load the model in `float16` for best quality.
-    - Falls back to **8-bit quantization** (`bitsandbytes`) if VRAM is insufficient.
-- **Audio Reconstruction**: Converts model masks to audio using STFT/ISTFT processing.
-
-## Local Development
-
-To run this application locally on your machine:
-
-1.  **Clone the repository:**
-    ```bash
-    git clone https://huggingface.co/spaces/lpeterl/sam-audio-webui
-    cd sam-audio-webui
-    ```
-
-2.  **Create a virtual environment (Recommended):**
-    ```bash
-    python3 -m venv venv
-    source venv/bin/activate
-    ```
-
-3.  **Install dependencies:**
-    ```bash
-    pip install -r requirements.txt
-    pip install gradio
-    ```
-
-4.  **Run the app:**
-    ```bash
-    python3 app.py
-    ```
-    *Note: `spaces` GPU decorators are mocked locally, so you don't need a ZeroGPU environment.*
 
 ## System Requirements
 
-- **VRAM**: ~21.6 GB for standard loading. ~12 GB with 8-bit quantization.
-- **Platform**: CUDA (NVIDIA GPU) required for quantization. Mac (MPS) supported for standard loading (requires high unified memory).
+- **VRAM**: ~21.6 GB for standard loading.
+- **Python**: >= 3.11 required by `perception-models` dependency.
 
 ## Acknowledgements
 

app.py

@@ -2,6 +2,17 @@ import gradio as gr
 import torch
 import torchaudio
 import tempfile
+
+try:
+    import spaces
+except ImportError:
+    class spaces:
+        @staticmethod
+        def GPU(duration=60):
+            def decorator(func):
+                return func
+            return decorator
+
 from sam_audio import SAMAudio, SAMAudioProcessor
 
 # Configuration
@@ -29,6 +40,7 @@ def save_audio(tensor, sample_rate):
         torchaudio.save(tmp.name, tensor, sample_rate)
         return tmp.name
 
+@spaces.GPU(duration=120)
 def separate_audio(audio_path, text_prompt):
     if not audio_path:
         return None, None
@@ -88,4 +100,4 @@ with gr.Blocks(title="SAM-Audio Demo") as demo:
     )
 
 # Launch
-demo.queue().launch(server_name="0.0.0.0", server_port=7860)
+demo.queue().launch()

requirements.txt

@@ -1,6 +1,9 @@
+gradio>=4.0.0
+torch>=2.0.0
+transformers>=4.38.0
+accelerate>=0.27.0
+scipy
+librosa
+spaces
 git+https://github.com/facebookresearch/sam-audio.git
-torch
 torchaudio
-gradio
-numpy
-scipy