changed gpu compute type

audio_processing.py

@@ -39,7 +39,7 @@ def process_audio(audio_file, translate=False, model_size="small"):
     try:
         device = "cuda" if torch.cuda.is_available() else "cpu"
         print(f"Using device: {device}")
-        compute_type = torch.bfloat16 if torch.cuda.is_available() else "float32"
+        compute_type = "int8" if torch.cuda.is_available() else "float32"
         audio = whisperx.load_audio(audio_file)
         model = whisperx.load_model(model_size, device, compute_type=compute_type)
 
@@ -53,7 +53,7 @@ def process_audio(audio_file, translate=False, model_size="small"):
         language_segments = []
         final_segments = []
         
-        overlap_duration = 2  # 2 seconds overlap
+        overlap_duration = OVERLAP  # 2 seconds overlap
         for i, chunk in enumerate(chunks):
             chunk_start_time = i * (CHUNK_LENGTH - overlap_duration)
             chunk_end_time = chunk_start_time + CHUNK_LENGTH