Forums

> Support for GPU models and installing new packages

Hi everyone,

Thanks to the organizers for creating this competition, this is a very interesting problem!

I have some questions:

- Does this competition support models running on GPUs, using pytorch for example?
- I know CodaLab supports different kinds of docker images: https://github.com/codalab/codalab-dockers, what is the docker image used on this competition?
- Is it possible to install any custom python packages? If not, what are the packages we can use?

Thanks,
Pedro

Posted by: prhb @ June 13, 2022, 8:33 p.m.

---

Hi Pedro,

Please find below the answers to your questions:

- Does this competition support models running on GPUs, using pytorch for example?
No, this competition is running on the default queue provided by Codalab, which does not feature a GPU.

- I know CodaLab supports different kinds of docker images: https://github.com/codalab/codalab-dockers, what is the docker image used on this competition?
Here is the link to the docker image used in this competition: https://hub.docker.com/layers/lc-challenge/didayolo/lc-challenge/latest/images/sha256-a8aec3412c3ecf1f6441d82a078e9c1f7d41d8e893d5201be33388e239989ada?context=explore

- Is it possible to install any custom python packages? If not, what are the packages we can use?
Yes, it’s possible to use other packages that were not provided in the docker image by simply including them in your zipped submission.

Best,
Hung

Posted by: hungnm @ June 14, 2022, 7:20 p.m.