Update app.py

app.py

@@ -14,13 +14,23 @@ from nemo.collections.asr.parts.utils.streaming_utils import FrameBatchMultiTask
 from nemo.collections.asr.parts.utils.transcribe_utils import get_buffered_pred_feat_multitaskAED
 
 
+model_id = "meta-llama/Meta-Llama-3-8B"
 
+llama_pipe = pipeline(
+    "text-generation", 
+    model=model_id, 
+    model_kwargs={"torch_dtype": torch.bfloat16}, 
+    device_map="auto",
+)
+
+
+'''
 torch.random.manual_seed(0)
 proc_model_name = "microsoft/Phi-3-mini-4k-instruct-gguf"
 proc_model = AutoModelForCausalLM.from_pretrained(proc_model_name)
 proc_model.to("cpu")
 proc_tokenizer = AutoTokenizer.from_pretrained(proc_model_name)
-
+'''
 
 
 
@@ -113,7 +123,8 @@ def generate_response(user_input):
     
 def CanaryPhi(audio_filepath):
     user_input = transcribe(audio_filepath)
-    response = generate_response(user_input)
+    #response = generate_response(user_input)
+    response = llama_pipe(user_input)
     return response