Skip to content
This repository has been archived by the owner on May 24, 2018. It is now read-only.

Error "Incorrect Device ID" on running kaggle_bowl #68

Closed
apeeyush opened this issue Mar 24, 2015 · 12 comments
Closed

Error "Incorrect Device ID" on running kaggle_bowl #68

apeeyush opened this issue Mar 24, 2015 · 12 comments

Comments

@apeeyush
Copy link

I tried running MNIST and kaggle_bowl example on an AWS GPU. I was able to successfully run the MNIST example but got an error message saying

Error "Incorrect Device ID" on running kaggle_bowl

on running kaggle_bowl.

However, I was able to run the kaggle_bowl example around a week ago!

Could you please help with where the problem might be?

@antinucleon
Copy link
Contributor

Please check your cuda configuration and dev setting. MNIST.conf is a CPU model.

@apeeyush
Copy link
Author

I ran the MNIST_CONV.conf file which is a GPU model. I am guessing it is some issue with OpenCV, however I am not sure!

@antinucleon
Copy link
Contributor

The error is not related to opencv. Check whether in the configuration there is two gpu selected by you only have one.

@apeeyush
Copy link
Author

I ran the code just after cloning from Github. It appears that the conf file contains 1 GPU by default.

@antinucleon
Copy link
Contributor

I will check it later today. The example is written before V2, so there may be some issues. Thanks

@apeeyush
Copy link
Author

I forgot to update the cloned repo (which I did last time when I successfully ran kaggle_bowl). Maybe this is causing the error. I'll try running the code after updating the repo.

@antinucleon
Copy link
Contributor

You need to use new mshadow from dmlc and new cxxnet.

@apeeyush
Copy link
Author

I tried to pull the latest code for mshadow and dmlc/cxxnet. However, when I run

./build.sh

The build completes but I get the following message at the end

/usr/bin/ld: skipping incompatible /usr/local/cuda/lib64/../lib/libcudart.so when searching for -lcudart

Could you please help with how I can build successfully. Thanks :)

@antinucleon
Copy link
Contributor

That's your cuda configuration problem. I am not able to help you in system admin area...

@apeeyush
Copy link
Author

I think the issue is because of some breaking changes in V2. I will check it in a few days and would be grateful if you could verify that the code works with the new version.

@tqchen
Copy link
Member

tqchen commented Mar 25, 2015

If you build completes without error, then it is OK. That message simply means you have 32bit cuda library in your searchpath and they skipped it

@tqchen tqchen closed this as completed Mar 25, 2015
@apeeyush
Copy link
Author

@tqchen The build completes without error. Also, I am able to run the MNIST_CONV.conf successfully. However, I am unable to run the kaggle bowl example. I am guessing that this might be because of some breaking changes in V2 since I was able to run the kaggle bowl example around a week ago.

@tqchen @antinucleon I would be grateful if you could verify that the example still works for you with the latest commits since I am unable to run them now.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants