update code in colab (#1)

- update code in colab (c3f3fd3f261f12d4bdc6f7db19c0f5a5ff56bafb)


Co-authored-by: XueHang <[email protected]>

README.md

@@ -15,7 +15,7 @@ widget:
 ---
 
 
-ChatYuan-large-v2是一个支持中英双语的功能型对话语言大模型(Paddle版本）。v2使用了和 v1版本相同的技术方案，在指令微调、人类反馈强化学习、思维链等方面进行了优化。
+ChatYuan-large-v2是一个支持中英双语的功能型对话语言大模型。v2使用了和 v1版本相同的技术方案，在指令微调、人类反馈强化学习、思维链等方面进行了优化。
 
 ChatYuan-large-v2 is a functional dialogue language model that supports bilingual Chinese and English. 
 ChatYuan-large-v2 uses the same technical solution as the v1 version, and has been optimized in terms of instruct-tuning, human feedback reinforcement learning and chain-of-thought.
@@ -23,7 +23,7 @@ ChatYuan-large-v2 uses the same technical solution as the v1 version, and has be
 <a href='https://huggingface.co/spaces/ClueAI/ChatYuan-large-v2' target="__blank">在线Demo</a> &nbsp; | 
   <a href='https://www.clueai.cn' target="__blank">使用API(large版)</a> &nbsp; | 
  &nbsp; <a href='https://github.com/clue-ai/ChatYuan' target="__blank">Github项目地址</a>&nbsp; |
-  &nbsp;<a href='https://colab.research.google.com/drive/1ZcLIJuemiojigrfjbsDMBWrX7JqXZX6I?usp=sharing' target="__blank">Colab在线试用</a> &nbsp; |
+  &nbsp;<a href='https://colab.research.google.com/drive/1JTSKy2HntPYHi6UvmUiwdHisFx1UrWlJ?usp=sharing' target="__blank">Colab在线试用</a> &nbsp; |
   &nbsp;<a href='https://mp.weixin.qq.com/s/FtXAnrhavA5u7hRyfm8j6Q' target="__blank">文章介绍</a> 
 
 
@@ -74,15 +74,11 @@ Based on the original functions of Chatyuan-large-v1, we optimized the model as
 加载模型：
  
  ```python
-# 加载模型
-from transformers import T5Tokenizer, T5ForConditionalGeneration
-tokenizer = T5Tokenizer.from_pretrained("ClueAI/ChatYuan-large-v2")
-model = T5ForConditionalGeneration.from_pretrained("ClueAI/ChatYuan-large-v2")
+# 加载模型,直接从paddlenlp中加载
+from paddlenlp.transformers import AutoTokenizer, T5ForConditionalGeneration
+tokenizer = AutoTokenizer.from_pretrained("ClueAI/ChatYuan-large-v2", from_hf_hub=False)
+model = T5ForConditionalGeneration.from_pretrained("ClueAI/ChatYuan-large-v2", from_hf_hub=False)
 # 该加载方式，在最大长度为512时 大约需要6G多显存
-# 如显存不够，可采用以下方式加载，进一步减少显存需求，约为3G
-# model = T5ForConditionalGeneration.from_pretrained("ClueAI/ChatYuan-large-v2").half()
-
-
 
  ```
 
@@ -90,10 +86,7 @@ model = T5ForConditionalGeneration.from_pretrained("ClueAI/ChatYuan-large-v2")
 ```python
 # 使用
 import torch
-from transformers import AutoTokenizer
 # 修改colab笔记本设置为gpu，推理更快
-device = torch.device('cuda')
-model.to(device)
 def preprocess(text):
   text = text.replace("\n", "\\n").replace("\t", "\\t")
   return text
@@ -105,12 +98,12 @@ def answer(text, sample=True, top_p=1, temperature=0.7):
   '''sample：是否抽样。生成任务，可以设置为True;
   top_p：0-1之间，生成的内容越多样'''
   text = preprocess(text)
-  encoding = tokenizer(text=[text], truncation=True, padding=True, max_length=512, return_tensors="pt").to(device) 
+  encoding = tokenizer(text=[text], truncation=True, padding=True, max_length=768, return_tensors="pd") 
   if not sample:
-    out = model.generate(**encoding, return_dict_in_generate=True, output_scores=False, max_new_tokens=512, num_beams=1, length_penalty=0.6)
+    out = model.generate(**encoding, return_dict_in_generate=True, output_scores=False, max_length=512, num_beams=1, length_penalty=0.4)
   else:
-    out = model.generate(**encoding, return_dict_in_generate=True, output_scores=False, max_new_tokens=512, do_sample=True, top_p=top_p, temperature=temperature, no_repeat_ngram_size=3)
-  out_text = tokenizer.batch_decode(out["sequences"], skip_special_tokens=True)
+    out = model.generate(**encoding, return_dict_in_generate=True, output_scores=False, max_length=512, do_sample=True, top_p=top_p, temperature=temperature, no_repeat_ngram_size=3)
+  out_text = tokenizer.batch_decode(out[0], skip_special_tokens=True)
   return postprocess(out_text[0])
 print("end...")
 ```