Building and running docker image on Ubuntu 20.04

[annalisasheehan]

Hello,

I've recently updated my computer to Ubuntu 20.04 and have since run into a lot of issues when running the SIMRDWN code and building the dockerfile.

After building the dockerfile and making YOLT3 i get the error: Makefile: 100 : recipe for target 'libdarknet.so' failed

When attempting to run the simrdwn.py i get the error: ./yolt3/darknet: error while loading shared libraries: libcudart.so.9.0: cannot open shared object file: No such file or directory

On my PC I have: Ubuntu 20.04 NVIDIA-SMI 455.32.00
CUDA Version: 11.1

Any help on why this may be happening would be greatly appreciated. Many Thanks, Annalisa

[sunyuxiu]

Hi.I do not build the docker when I run the script.For my reason,I think it is not necessary to build to the docker because the docker just is a tool to make the script more convenient run.In my environment,I build a anaconda env tlist to solve the environment.And now,I meet a new error,maybe we can communicate each other when you run the test file.I am looking forward your letter.

[sunyuxiu]

Hi.I do not build the docker when I run the script.For my reason,I think it is not necessary to build to the docker because the docker just is a tool to make the script more convenient run.In my environment,I build a anaconda env tlist to solve the environment.And now,I meet a new error,maybe we can communicate each other when you run the test file.I am looking forward your letter.

Maybe this issues can help you: https://github.com/avanetten/simrdwn/issues/81

[annalisasheehan]

Hi Sunyuxiu,

Thank you for your comment and the link to the other issue. I've had a look at this and also thought that it may be due to the CUDA version however there's limited information that i can find on how to solve the issue and whether i can change CUDA to a previous version with Ubuntu 20.04. I was perfectly able to run the docker image and code on Ubuntu 18.04 and unsure why updating to Ubuntu 20.04 has had such a negative impact on docker/cuda etc. Any help on this would be greatly appreciated?

I have also tried building a conda environment (based on the dockerfile) however found that there are hardcoded paths in the code which assume you are working in a docker image. Is that the issue you have also found?

[sunyuxiu]

Hi.Nice to receive your reply!I do not know what the hardcode means in your last reply.And,I only use conda environment instead of docker environment.

Hi Sunyuxiu,

Thank you for your comment and the link to the other issue. I've had a look at this and also thought that it may be due to the CUDA version however there's limited information that i can find on how to solve the issue and whether i can change CUDA to a previous version with Ubuntu 20.04. I was perfectly able to run the docker image and code on Ubuntu 18.04 and unsure why updating to Ubuntu 20.04 has had such a negative impact on docker/cuda etc. Any help on this would be greatly appreciated?

I have also tried building a conda environment (based on the dockerfile) however found that there are hardcoded paths in the code which assume you are working in a docker image. Is that the issue you have also found?

[sunyuxiu]

In addition,what stage do you have run?Training or testing?

[sunyuxiu]

Hi,do you meet the similar question like this when you test the net?https://github.com/avanetten/simrdwn/issues/101