Closed SmashChen closed 1 year ago
It looks to me like nvidia recently took down a lot of older cuda versions from their docker hub. I don't know if it is possible to still download them somehow. For this repository I updated the cuda version to 11.3.1
FROM nvidia/cuda:11.3.1-cudnn8-devel-ubuntu16.04
You can check the available versions here.
I am using cuda 11.3 for pytorch in this repository as well. You might also need to adjust the pytorch, mmdet cuda versions etc. for things to work out with the transfuser repository.
Thanks for your reply, your suggestion works. I also want to ask another question:
When I submitted our team's docker, I was told that our team did not have permission from the host.
How should I get the permission? I have asked the staff but they have been slow to respond. I hope you can give me some advice, thank you very much.
You just need to make a team, apply for the benchmark and then wait. The team is manually approved by the staff. I don't think there is any way to speed things up if you already asked them.
Ok, i get it. Sincerely thank you for your reply.
Hi authors, Thank you for your contribution! I noticed that you are also performing Carla Leaderboard benchmark recently. Congratulations on refreshing your score again. I have a question about docker configuration and would like to ask you: When I run "make_docker.sh" using the "Dockerfile.master" you provided on T-PAMI 22, it throws an error: ERROR: failed to solve: nvidia/cuda:10.2-cudnn7-devel-ubuntu16.04: docker .io/nvidia/cuda:10.2-cudnn7-devel-ubuntu16.04: not found. I think it's because the Ubuntu16.04 image has been removed, but I can't find a solution.