Update app.py

app.py

@@ -21,7 +21,7 @@ proc_model = AutoModelForCausalLM.from_pretrained(
     torch_dtype=torch.float16,
     revision="653ee820c4f2ee66427e997b4a8ca3e9323e7d46",
     trust_remote_code=True,
-    low_cpu_mem_usage=True,
+    #low_cpu_mem_usage=True,
 )
 
 proc_model.to("cpu")
@@ -105,8 +105,6 @@ def transcribe(audio_filepath):
 
 start = {"role": "system", "content": "You are a helpful digital assistant. Please provide safe, ethical and accurate information to the user."}
 
-
-
 def generate_response(user_input):  
     messages = [start, {"role": "user", "content": user_input}]
     inputs = proc_tokenizer.apply_chat_template(
@@ -135,7 +133,7 @@ def CanaryPhiVits(user_voice):
     response = generate_response(user_input)
     print(response)
     if response.startswith(user_input):
-        response = chatty_response.replace(user_input, '', 1)
+        response = response.replace(user_input, '', 1)
     print(response)
     chatty_response = text_to_speech(response)