解决pip install icetk运行ChatGLM-6B的问题

出现的问题

下载THUDM/chatglm-6b的代码，运行6b模型的时候，出现了pip install icetk的ImportError问题。
下面是在ptuning文件夹下运行main.py报错信息：

Import Error: This modeling file requires the following packages that were not fouond in your environment:icetk. Run pip install icetk

于是单独执行一行代码self.tokenizer = AutoTokenizer.from_pretrained(""THUDM/chatglm-6b, trust_remote_code=True)的时候也出现了错误，报错信息：

ImportError: This modeling file requires the following packages that were not found in your environment: icetk. Run pip install icetk

眼疾手快，反手就是pip install icetk，如果这么顺利的话，就不会有这篇水文了。容我娓娓道来。

我的环境

• 显卡4090
• Build cuda_11.7.r11.7/compiler.31442593_0
• torch ：1.13.1 cu117

直接执行pip install icetk后，信息如下：

PS D:\nlp-research> pip install icetk
Looking in indexes: https://mirrors.aliyun.com/pypi/simple/
Collecting icetk
  Downloading https://mirrors.aliyun.com/pypi/packages/bf/8a/731927e0901273815b779e6ce0e081a95ecf78835ff80be30830505ae06c/icetk-0.0.7-py3-none-any.whl (16 kB)
Requirement already satisfied: tqdm in d:\apps\python39\lib\site-packages (from icetk) (4.65.0)
Requirement already satisfied: requests in d:\apps\python39\lib\site-packages (from icetk) (2.31.0)
Collecting torchvision (from icetk)
  Downloading https://mirrors.aliyun.com/pypi/packages/d8/48/e2a056436033da54856d793e12dc0fcf8cdd179fd4cd0d1ce7c7ce659797/torchvision-0.15.2-cp39-cp39-win_amd64.whl (1.2
 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 925.5 kB/s eta 0:00:00
Requirement already satisfied: sentencepiece in d:\apps\python39\lib\site-packages (from icetk) (0.1.99)
Requirement already satisfied: filelock in d:\apps\python39\lib\site-packages (from icetk) (3.12.2)
Collecting protobuf<3.19 (from icetk)
  Downloading https://mirrors.aliyun.com/pypi/packages/cc/57/4eefdf3bb61a997532a6022d2a9562b9c3acb0287b60f27d5a872acf29cd/protobuf-3.18.3-cp39-cp39-win_amd64.whl (912 kB
)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 912.5/912.5 kB 916.9 kB/s eta 0:00:00
Requirement already satisfied: charset-normalizer<4,>=2 in d:\apps\python39\lib\site-packages (from requests->icetk) (3.1.0)
Requirement already satisfied: idna<4,>=2.5 in d:\apps\python39\lib\site-packages (from requests->icetk) (3.4)
Requirement already satisfied: urllib3<3,>=1.21.1 in d:\apps\python39\lib\site-packages (from requests->icetk) (2.0.3)
Requirement already satisfied: certifi>=2017.4.17 in d:\apps\python39\lib\site-packages (from requests->icetk) (2023.5.7)
Requirement already satisfied: numpy in d:\apps\python39\lib\site-packages (from torchvision->icetk) (1.25.0)
Collecting torch==2.0.1 (from torchvision->icetk)
  Downloading https://mirrors.aliyun.com/pypi/packages/48/f4/d0b61525a3d3db78636f1937d1bc24cbb39abc57484a545b72b6ab35c114/torch-2.0.1-cp39-cp39-win_amd64.whl (172.4 MB) 
     ╸━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.4/172.4 MB 948.8 kB/s eta 0:03:00

分析

上面是我直接安装icetk遇到的信息，本来我的环境就有torch:1.13.1 cu117了，它还在安装torch-2.0.1-cp39-cp39-win_amd64.whl ，经分析我的环境中没有torchvision，安装icetk的时候需要依赖torchvision，于是就安装了torchvision-0.15.2，跟我当前torch1.13.1 cu117不兼容。

Collecting torch==2.0.1 (from torchvision->icetk)
Downloading https://mirrors.aliyun.com/pypi/packages/48/f4/d0b61525a3d3db78636f1937d1bc24cbb39abc57484a545b72b6ab35c114/torch-2.0.1-cp39-cp39-win_amd64.whl (172.4 MB)

如果torch==2.0 是CPU不是GPU的，还要折腾一番，于是索性手动安装torchvision了，一个个尝试安装，直到安装icetk的时候不会再安装torch==2.0.1就好了。

解决步骤

1. 随便尝试 pip install torchvision==1.2.3，出现如下：

   C:\Users\Administrator>pip install torchvision1.2.3
   Looking in indexes: https://mirrors.aliyun.com/pypi/simple/
   ERROR: Could not find a version that satisfies the requirement torchvision1.2.3 (from versions: 0.1.6, 0.1.7, 0.1.8, 0.1.9, 0.2.0, 0.2.1, 0.2.2, 0.2.2.post2, 0.2.2.post3, 0.9.0, 0.9.1, 0.10.0, 0.10.1, 0.11.0, 0.11.1, 0.11.2, 0.11.3, 0.12.0, 0.13.0, 0.13.1, 0.14.0, 0.14.1, 0.15.0, 0.15.1, 0.15.2)
   ERROR: No matching distribution found for torchvision==1.2.3

   可以看出分析那里看到的信息，安装的是最新版本的torchvision，难怪会依赖torch==2.0

2. 继续尝试·pip install torchvision==0.15.0·，出现信息如下：
   

   依然在下载torch==2.0的版本。

3. 尝试pip install torchvision==0.14.1
   

   可以看到截图，一马平川了。成功安装icetk。

这篇好文章是转载于：学新通技术网