search

ziansu / prorec

Official Implementation of NeurIPS 2024 paper - Source Code Foundation Models are Transferable Binary Analysis Knowledge Bases
3 stars 0 forks source link

problem #3

Open zzy443320 opened 13 hours ago

zzy443320 commented 13 hours ago

I encountered the following problem when running the binsum.ipynb file. After looking up relevant information, I thought it was an issue with the version of huggingface_hub that caused the DownloadConfig class to not have the use_auth_token attribute. However, I've tried several versions and this problem still persists. Could you please tell me how to solve it? image

ziansu commented 12 hours ago

I tried with the latest huggingface_hub==0.26.2 in a new environment and it works well. Maybe you can try with this version.

zzy443320 commented 12 hours ago

我在新环境中尝试了最新的，效果很好。也许你可以试试这个版本。huggingface_hub==0.26.2

1、我重新下载了0.26.2版本但还是出现了相同的问题，除了考虑huggingface_hug版本问题，还有其他解决办法吗 2、实验中用到的软件包以及库是否都需要相同的的channel？如果是的话，需要哪种channel呢？ 3、如果我想要最小复现实验，是否按照ipynb文件中的单元格顺序来执行代码即可？

ziansu commented 4 hours ago
  1. 看起来是evaluate的版本问题，详见https://github.com/huggingface/evaluate/issues/638
  2. channel应该不是特别重要
  3. 最小复现实验只需要跑ipynb
zzy443320 commented 4 hours ago
  1. 看起来是evaluate的版本问题，详见Evaluate fails to load all metrics. huggingface/evaluate#638
  2. channel应该不是特别重要
  3. 最小复现实验只需要跑ipynb

1、我在download_config.py文件中给DownloadConfig类增加了use_auth_token属性就不再出现错误，这种方法会对程序其他地方有影响吗？ image 2、我在运行4.1单元格的代码时出现了卡顿，是因为数据过大导致的吗？ image 出现红×是因为我中断了运行，而不是出现错误

ziansu commented 3 hours ago
  1. 能下载就行
  2. 应该就是单纯的慢，因为是串行访问OpenAI API
zzy443320 commented 2 hours ago
  1. 能下载就行
  2. 应该就是单纯的慢，因为是串行访问OpenAI API

非常感谢您的帮助。我还有最后一个问题，我已经成功运行了整个binsum.ipynb文件并得到了json文件，我还需要对json文件进行什么操作吗