Update app.py

app.py

@@ -1,9 +1,10 @@
 import torch
 import torchaudio
 from einops import rearrange
-
 import gradio as gr
 import spaces
+import os
+import uuid
 
 # Importing the model-related functions
 from stable_audio_tools import get_pretrained_model
@@ -13,9 +14,14 @@ from stable_audio_tools.inference.generation import generate_diffusion_cond
 @spaces.GPU(duration=120)  # Allocate GPU only when this function is called
 def generate_audio(prompt, seconds_total=30, steps=100, cfg_scale=7):
     device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    # Fetch the Hugging Face token from the environment variable
+    hf_token = os.getenv('HF_TOKEN')
+    if not hf_token:
+        raise EnvironmentError("HF_TOKEN environment variable not set")
     
     # Download and set up the model
-    model, model_config = get_pretrained_model("stabilityai/stable-audio-open-1.0")
+    model, model_config = get_pretrained_model("stabilityai/stable-audio-open-1.0", use_auth_token=hf_token)
     sample_rate = model_config["sample_rate"]
     sample_size = model_config["sample_size"]
 
@@ -46,12 +52,14 @@ def generate_audio(prompt, seconds_total=30, steps=100, cfg_scale=7):
 
     # Peak normalize, clip, convert to int16
     output = output.to(torch.float32).div(torch.max(torch.abs(output))).clamp(-1, 1).mul(32767).to(torch.int16).cpu()
-    
+
+    # Generate a unique filename for the output
+    unique_filename = f"output_{uuid.uuid4().hex}.wav"
     # Save to file
-    torchaudio.save("output.wav", output, sample_rate)
+    torchaudio.save(unique_filename, output, sample_rate)
 
     # Return the path to the generated audio file
-    return "output.wav"
+    return unique_filename
 
 # Setting up the Gradio Interface
 interface = gr.Interface(