Refactor to use upstream automodel.

app.py

@@ -1,28 +1,9 @@
-# -*- coding: utf-8 -*-
-"""Fujisaki_CPU.ipynb
-
-Automatically generated by Colaboratory.
-
-Original file is located at
-    https://colab.research.google.com/drive/1Damnr0Ha4zZAlKFvne9cu76uuElLNYus
-
-李萌萌的电子骨灰盒
-----
-
-这是一个通过ChatGLM模型训练的李萌萌的数字分身，你可以在问题栏目填入内容，或者什么都不填，来观察李萌萌到底会说些什么。
-T4级别的GPU已经可以很胜任这个任务了。
-
-### 安装依赖
-"""
-
-#from modeling_chatglm import ChatGLMForConditionalGeneration
 import torch
-import sys
 
 from transformers import AutoTokenizer, GenerationConfig, AutoModel
 
-model = AutoModel.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True).float()
-tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True)
+model = AutoModel.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True, revision="fdb7a60").float()
+tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True, revision="fdb7a60")
 
 from peft import PeftModel
 peft_path = 'ljsabc/Fujisaki_GLM'      # change it to your own
@@ -36,6 +17,7 @@ model = PeftModel.from_pretrained(
 print(model.peft_config)
 # We have to use full precision, as some tokens are >65535
 model.eval()
+print(model)
 
 torch.set_default_tensor_type(torch.FloatTensor)
 def evaluate(context, temperature, top_p, top_k):