THUDM / CogDL

CogDL: A Comprehensive Library for Graph Deep Learning (WWW 2023)

Home Page:https://cogdl.ai

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

导入CogDL及其子模块的时间非常长/It takes a long time to import CogDL and its sub-modules

FrancisGrace opened this issue · comments

🐛 Bug

执行导入CogDL及其子模块语句的时间非常长。
It takes a long time to import CogDL and its sub-modules.

To Reproduce

Steps to reproduce the behavior:

import cogdl
print('Hello World!')

运行此代码后五分钟内未输出“Hello World!”。
After running the code, 'Hello World!' are not printed in 5 minutes.

Expected behavior

运行此代码后立刻输出“Hello World!”。
After running the code, 'Hello World!' should be printed immediately.

Environment

  • CogDL version: 0.5.3(使用v0.5.0正常/behaves normally in v0.5.0)
  • OS (e.g., Linux): Windows 10
  • Python version: 3.8.13
  • PyTorch version: 1.11.0
  • CUDA/cuDNN version (if applicable): CUDA 11.1
  • Any other relevant information: None

Additional context

Hi @FrancisGrace,

我猜测是在编译cogdl的稀疏算子,能否看一下~/.cache/torch_extensions里的情况。

第一次导入的时候需要对底层算子进行编译,所以需要花费几分钟。后续导入的话应该就会比较快。

🐛 Bug

执行导入CogDL及其子模块语句的时间非常长。 It takes a long time to import CogDL and its sub-modules.

To Reproduce

Steps to reproduce the behavior:

import cogdl
print('Hello World!')

运行此代码后五分钟内未输出“Hello World!”。 After running the code, 'Hello World!' are not printed in 5 minutes.

Expected behavior

运行此代码后立刻输出“Hello World!”。 After running the code, 'Hello World!' should be printed immediately.

Environment

  • CogDL version: 0.5.3(使用v0.5.0正常/behaves normally in v0.5.0)
  • OS (e.g., Linux): Windows 10
  • Python version: 3.8.13
  • PyTorch version: 1.11.0
  • CUDA/cuDNN version (if applicable): CUDA 11.1
  • Any other relevant information: None

Additional context

我也遇到了此问题,请问你解决了吗?

是重装cogdl之后出现的这种情况吗

并不是,我前段时间用的时候还没问题,有一段时间没用之后再运行导入时间就变得很长,我现在将cogdl改成了0.5.0版本就可以快速运行了

我这里也是不知道如何解决。一直卡在导入中