昇思25天学习打卡营第八天|应用实践/LLM原理与实践/MindNLP ChatGLM-6B StreamChat

心得

特别注意：

cell 1 的原始代码

%%capture captured_output
# 实验环境已经预装了mindspore==2.2.14，如需更换mindspore版本，可更改下面mindspore的版本号
!pip uninstall mindspore -y
!pip install -i https://pypi.mirrors.ustc.edu.cn/simple mindspore==2.2.14
 

要在这个代码最后，添加

!pip install mindnlp
!pip install mdtex2html

否则会报错

添加代码后正常。

打卡截图：

MindNLP ChatGLM-6B StreamChat

本案例基于MindNLP和ChatGLM-6B实现一个聊天应用。

1 环境配置

[1]:

%%capture captured_output

# 实验环境已经预装了mindspore==2.2.14，如需更换mindspore版本，可更改下面mindspore的版本号

!pip uninstall mindspore -y

!pip install -i https://pypi.mirrors.ustc.edu.cn/simple mindspore==2.2.14

!pip install mindnlp

!pip install mdtex2html

配置网络线路

[2]:

!export HF_ENDPOINT=https://hf-mirror.com

2 代码开发

下载权重大约需要10分钟

[3]:

from mindnlp.transformers import AutoModelForSeq2SeqLM, AutoTokenizer

import gradio as gr

import mdtex2html

​

model = AutoModelForSeq2SeqLM.from_pretrained('ZhipuAI/ChatGLM-6B', mirror="modelscope").half()

model.set_train(False)

tokenizer = AutoTokenizer.from_pretrained('ZhipuAI/ChatGLM-6B', mirror="modelscope")

Building prefix dict from the default dictionary ...
Dumping model to file cache /tmp/jieba.cache
Loading model cost 1.034 seconds.
Prefix dict has been built successfully.

100%

 773/773 [00:00<00:00, 53.8kB/s]

100%

 32.6k/32.6k [00:00<00:00, 2.95MB/s]

Downloading shards: 100%

 8/8 [16:37<00:00, 107.48s/it]

100%

 1.62G/1.62G [02:07<00:00, 14.4MB/s]

100%

 1.75G/1.75G [02:19<00:00, 10.2MB/s]

100%

 1.84G/1.84G [02:27<00:00, 20.0MB/s]

100%

 1.78G/1.78G [02:22<00:00, 10.4MB/s]

100%

 1.75G/1.75G [02:23<00:00, 12.3MB/s]

100%

 1.75G/1.75G [02:18<00:00, 14.5MB/s]

100%

 1.00G/1.00G [01:18<00:00, 18.0MB/s]

100%

 1.00G/1.00G [01:19<00:00, 12.8MB/s]

Loading checkpoint shards: 100%

 8/8 [00:44<00:00,  4.84s/it]

100%

 441/441 [00:00<00:00, 43.9kB/s]

100%

 2.58M/2.58M [00:00<00:00, 20.4MB/s]

可以修改下列参数和prompt体验模型

[5]:

prompt = '你好'

history = []

response, _ = model.chat(tokenizer, prompt, history=history, max_length=20)

response

[5]:

'你好👋！我是人工智能助手 ChatGLM-6B'

[6]:

你好

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[6], line 1
----> 1 你好

NameError: name '你好' is not defined

[7]:

prompt = '你好'

[8]:

prompt = '今天好热'

history = []

response, _ = model.chat(tokenizer, prompt, history=history, max_length=20)

response

-

[8]:

'是的，今天天气很热。夏季通常是气温较高、阳光强烈的季节'

[9]:

import time

print(time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()),'guojun0718')

2024-07-18 08:44:17 guojun0718

[ ]:

​