Update app.py

app.py

@@ -41,6 +41,8 @@ frame_asr = FrameBatchMultiTaskAED(
 
 amp_dtype = torch.float16
 
+llm_pipeline = transformers.pipeline("text-generation", model="meta-llama/Meta-Llama-3-8B", model_kwargs={"torch_dtype": torch.bfloat16}, device_map="auto")
+
 def convert_audio(audio_filepath, tmpdir, utt_id):
 	"""
 	Convert all files to monochannel 16 kHz wav files.
@@ -222,12 +224,6 @@ def on_src_or_tgt_lang_change(src_lang_value, tgt_lang_value, pnc_value):
 		)
 	return src_lang, tgt_lang, pnc
 
-llm_pipeline = transformers.pipeline(
-    "text-generation", 
-    model="meta-llama/Meta-Llama-3-8B", 
-    model_kwargs={"torch_dtype": torch.bfloat16}, 
-    device_map="auto"
-)
 
 with gr.Blocks(
 	title="NeMo Canary Model",
@@ -271,6 +267,7 @@ with gr.Blocks(
 				)
 
 		with gr.Column():
+            
 			gr.HTML("<p>Run the model.</p>")
 
 			go_button = gr.Button(
@@ -313,7 +310,5 @@ with gr.Blocks(
 	)
     
     
-
-
 demo.queue()
 demo.launch(share=True)