Update app.py

app.py

@@ -7,11 +7,11 @@ from llama_cpp import Llama
 from huggingface_hub import hf_hub_download #load from huggingfaces 
 
 
-llm = Llama(model_path= hf_hub_download(repo_id="TheBloke/OpenAssistant-Llama2-13B-Orca-v2-8K-3166-GGML", filename="openassistant-llama2-13b-orca-v2-8k-3166.ggmlv3.q6_K.bin"), n_ctx=2048) #download model from hf/ n_ctx=2048 for high ccontext length
+llm = Llama(model_path= hf_hub_download(repo_id="TheBloke/Dolphin-Llama2-7B-GGML", filename="dolphin-llama2-7b.ggmlv3.q4_1.bin"), n_ctx=2048) #download model from hf/ n_ctx=2048 for high ccontext length
 
 history = []
 
-pre_prompt = " The user and the AI are having a conversation : "
+pre_prompt = " The user and the AI are having a conversation : <|endoftext|> \n "
 
 def generate_text(input_text, history):
     print("history ",history)
@@ -23,7 +23,7 @@ def generate_text(input_text, history):
         input_text_with_history = f"{history[-1][1]}"+ "\n"
         input_text_with_history += f"<|prompter|> {input_text}" + "\n" +" <|assistant|>"
     print("new input", input_text_with_history)
-    output = llm(input_text_with_history, max_tokens=1024, stop=["<|prompter|>", "<|endoftext|>"], stream=True)
+    output = llm(input_text_with_history, max_tokens=1024, stop=["<|prompter|>", "<|endoftext|>", "<|endoftext|> \n"], stream=True)
     for out in output:
      stream = copy.deepcopy(out)
      print(stream["choices"][0]["text"])