Update app.py

app.py

@@ -8,6 +8,7 @@ from models import AudioDiffusion, DDPMScheduler
 from audioldm.audio.stft import TacotronSTFT
 from audioldm.variational_autoencoder import AutoencoderKL
 from gradio import Markdown
+import spaces
 
 class Tango:
     def __init__(self, name="declare-lab/tango", device="cuda:0"):
@@ -67,11 +68,13 @@ class Tango:
             return list(self.chunks(outputs, samples))
 
 # Initialize TANGO
-if torch.cuda.is_available():
-    tango = Tango()
-else:
-    tango = Tango(device="cpu")
 
+tango = Tango(device="cpu")
+tango.vae.to("cuda")
+tango.stft.to("cuda")
+tango.model.to("cuda")
+    
+@spaces.GPU(duration=60)
 def gradio_generate(prompt, steps, guidance):
     output_wave = tango.generate(prompt, steps, guidance)
     # output_filename = f"{prompt.replace(' ', '_')}_{steps}_{guidance}"[:250] + ".wav"
@@ -101,8 +104,8 @@ def gradio_generate(prompt, steps, guidance):
 # """
 description_text = ""
 # Gradio input and output components
-input_text = gr.inputs.Textbox(lines=2, label="Prompt")
-output_audio = gr.outputs.Audio(label="Generated Audio", type="filepath")
+input_text = gr.Textbox(lines=2, label="Prompt")
+output_audio = gr.Audio(label="Generated Audio", type="filepath")
 denoising_steps = gr.Slider(minimum=100, maximum=200, value=100, step=1, label="Steps", interactive=True)
 guidance_scale = gr.Slider(minimum=1, maximum=10, value=3, step=0.1, label="Guidance Scale", interactive=True)
 
@@ -133,8 +136,8 @@ gr_interface = gr.Interface(
         ["A helicopter is in flight"],
         ["A dog barking and a man talking and a racing car passes by"],
     ],
-    cache_examples=False, # Turn on to cache.
+    cache_examples="lazy", # Turn on to cache.
 )
 
 # Launch Gradio app
-gr_interface.launch()
+gr_interface.queue(10).launch()