Update app.py

app.py

@@ -15,13 +15,27 @@ vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
 speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
 
+# Quantize the models
+def quantize_model(model):
+    quantized_model = torch.quantization.quantize_dynamic(
+        model, {torch.nn.Linear}, dtype=torch.qint8
+    )
+    return quantized_model
+
+model = quantize_model(model)
+vocoder = quantize_model(vocoder)
+
+# JIT compile the models for faster inference
+model = torch.jit.script(model)
+vocoder = torch.jit.script(vocoder)
+
+# Use inference mode for faster computation
+@torch.inference_mode()
 def text_to_speech(text):
     inputs = processor(text=text, return_tensors="pt")
     speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
-    
     output_path = "output.wav"
     sf.write(output_path, speech.numpy(), samplerate=16000)
-    
     return output_path
 
 # Create Gradio interface
@@ -30,7 +44,7 @@ iface = gr.Interface(
     inputs=gr.Textbox(label="Enter text to convert to speech"),
     outputs=gr.Audio(label="Generated Speech"),
     title="Text-to-Speech Converter",
-    description="Convert text to speech using the SpeechT5 model."
+    description="Convert text to speech using the optimized SpeechT5 model."
 )
 
 # Launch the app