bug fix

app.py

@@ -2,21 +2,22 @@ import os
 import gradio as gr
 import torch
 import torchaudio
-from peft import (
+from peft.src.peft import (
     LoraConfig,
     get_peft_model,
     get_peft_model_state_dict,
     prepare_model_for_int8_training,
     set_peft_model_state_dict,
 )
-from transformers import GenerationConfig, LlamaForCausalLM, LlamaTokenizer, LlamaConfig
+from hf.transformers.src.transformers.generation import GenerationConfig
+from hf.transformers.src.transformers.models.llama import LlamaForCausalLM, LlamaTokenizer, LlamaConfig
 from utils.prompter import Prompter
 import datetime
 import time,json
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-base_model = "./Llama-2-7b-chat-hf-qformer/"
+base_model = "Llama-2-7b-chat-hf-qformer/"
 
 prompter = Prompter('alpaca_short')
 tokenizer = LlamaTokenizer.from_pretrained(base_model)