search

megagonlabs / sudowoodo

The source code of the Sudowoodo paper in ICDE 2023
BSD 3-Clause "New" or "Revised" License
14 stars 6 forks source link

Issue running column type detection code #1

Open ysunbp opened 1 year ago

ysunbp commented 1 year ago

Hi! Thank you for your hard work! I tried to run through the code of the column type detection part, yet encountered the following:

image

I wonder if there are any problems with my installation. The environment I use: torch 1.9.0+cu111 python 3.7.10 CUDA 11.1

RunhuiWang commented 1 year ago

Did you install Apex? Are you running on multiple GPUs?

ysunbp commented 1 year ago

Hi! I am running on single GPU. I specify the GPU usage in the python file as: os.environ['CUDA_VISIBLE_DEVICES'] = '0'

RunhuiWang commented 1 year ago

What GPUs are you using? What is your OS? Please provide more details.

ysunbp commented 1 year ago

Hi I have installed Apex following your instructions. The GPU I used is NVIDIA GeForce 3090. The OS is Linux.

RunhuiWang commented 1 year ago

It looks like a software&hardware compatibility issue. Please try the solution in this post https://discuss.pytorch.org/t/rtx-3090-compatibility/110014

GaryZ055 commented 1 year ago

Ran into the same issue with Tesla-V100. The reason seems like the compatibility issue between torch 1.9.0 and current version of Apex. @RunhuiWang which version of Apex did you use when running the experiment?

RunhuiWang commented 1 year ago

Ran into the same issue with Tesla-V100. The reason seems like the compatibility issue between torch 1.9.0 and current version of Apex. @RunhuiWang which version of Apex did you use when running the experiment?

The Apex version was 0.1. Have you tried newer version(s) of torch? Based on my experience, newer versions of torch also work for this project.