change to GPU

app.py

@@ -14,7 +14,7 @@ from utils import get_sentence, tts_interface
 # "cuda" - NVIDIA GPU
 # "cpu" - Plain CPU
 # "mps" - Apple silicon
-device = "cpu"
+device = "cuda"
  
 # Load Mistral LLM
 print("Loading Mistral LLM")
@@ -24,7 +24,7 @@ mistral_llm = Llama(model_path=mistral_model_path,n_gpu_layers=35,max_new_tokens
 
 # Load Whisper ASR model
 print("Loading Whisper ASR")
-whisper_model = WhisperModel("large-v3", device="cpu", compute_type="float32")
+whisper_model = WhisperModel("large-v3", device="cuda", compute_type="float16")
 
 
 # Get all available voices from edge_tts