update cuda line code

app.py

@@ -13,6 +13,8 @@ import spaces
 load_dotenv()
 default_gemini_api_key = os.getenv('gemini_api_key')
 
+device = 0 if torch.cuda.is_available() else "cpu"
+
 def configure_genai(api_key, model_variant):
     genai.configure(api_key=api_key)
     return genai.GenerativeModel(model_variant)
@@ -73,7 +75,6 @@ def transcribe(youtube_url, audio_file, whisper_model, gemini_api_key, gemini_pr
             gemini_api_key = default_gemini_api_key
         model = configure_genai(gemini_api_key, gemini_model_variant)
 
-        device = 0 if torch.cuda.is_available() else "cpu"
         pipe = pipeline(
             task="automatic-speech-recognition",
             model=whisper_model,
@@ -163,9 +164,9 @@ with gr.Blocks(theme='NoCrypt/miku') as demo:
         with gr.Column():
             input_type = gr.Radio(["YouTube URL", "Audio File"], label="Input Type", value="YouTube URL", info="Please consider using the audio file if you face any issues with the YouTube URL. Currently youtube is banning HuggingFace IP Addresses.")
             with gr.Row():
-                youtube_url = gr.Textbox(label="YouTube URL", info="Input the full URL of the YouTube video you want to transcribe and summarize. Example: https://www.youtube.com/watch?v=VIDEO_ID")
-                audio_file = gr.File(label="Upload Audio File", visible=False, file_types=['.wav', '.flac', '.mp3'])
-                whisper_model = gr.Dropdown(["openai/whisper-tiny", "openai/whisper-base", "openai/whisper-small", "openai/whisper-medium", "openai/whisper-large-v3"], label="Whisper Model", value="openai/whisper-large-v3", info="Tiny is the fastest model, but it's not the best quality. large-v3 is the best quality, but it's the slowest model.")
+                youtube_url = gr.Textbox(label="YouTube URL", visible=False, info="Input the full URL of the YouTube video you want to transcribe and summarize. Example: https://www.youtube.com/watch?v=VIDEO_ID")
+                audio_file = gr.File(label="Upload Audio File", visible=True, file_types=['.wav', '.flac', '.mp3'])
+                whisper_model = gr.Dropdown(["openai/whisper-tiny", "openai/whisper-base", "openai/whisper-small", "openai/whisper-medium", "openai/whisper-large-v3", "distil-whisper/distil-large-v3"], label="Whisper Model", value="distil-whisper/distil-large-v3", info="Tiny is the fastest model, but it's not the best quality. large-v3 is the best quality, but it's the slowest model.")
                 gemini_model_variant = gr.Dropdown(["gemini-1.5-flash", "gemini-1.5-pro"], label="Gemini Model Variant", value="gemini-1.5-pro", info="Gemini-1.5-flash is the fastest model, but it's not the best quality. Gemini-1.5-pro is the best quality, but it's slower")
             define_language = gr.Checkbox(label="Define Language", value=False, info="If you want to define the language, check this box")
             language = gr.Dropdown(["id","en", "es", "fr", "de", "it", "pt", "ru", "ja", "ko", "zh"], label="Language", value=None, info="Select the language for transcription", visible=False)