Closed Ge-Lab closed 6 months ago
@Ge-Lab if possible, please structure your issue using code fences. See this guide for details. It will make it easier to read and understand if you place logs inside code blocks, for example.
This guide suggests a few things to try. Interestingly, it apperas you should set CUDA_VISIBLE_DEVICES=0
within the container itself, but SINGULARITYENV_CUDA_VISIBLE_DEVICES=0
outside of the container. Were you setting the variable appropriately, within the container or outside of it?
Additionally, since you have two GPUs you will want to set this variable to 0,1.
Do you have the nvidia-container-cli
installed as suggested on the support page?
Hi Daniele,
Thanks for your response. I updated the post as you suggested.
How can I set CUDA_VISIBLE_DEVICES=0
within the container and SINGULARITYENV_CUDA_VISIBLE_DEVICES=0
outside of the container? I just typed export CUDA_VISIBLE_DEVICES=0
in the terminal and then typed the apptainer command to run.
I do have nvidia-container-cli
.
@Ge-Lab first lets try setting export SINGULARITYENV_CUDA_VISIBLE_DEVICES=0
before you run the container since that is easier. Give that a try and let me know if you are still having issues.
Hi @Ge-Lab , this has been inactive for more than a month, so I'll close. But please feel free to reopen with more details now if you want to follow up!
Hi,
I followed the instructions on deepvariant quick start (https://github.com/google/deepvariant/blob/r1.6/docs/deepvariant-quick-start.md) to create deepvariant_1.6.0.sif and deepvariant_1.6.0-gpu.sif successfully using apptainer.
Then, I followed the complete genomics T7 case study (https://github.com/google/deepvariant/blob/r1.6/docs/deepvariant-complete-t7-case-study.md) to have some test runs.
CPU version I run the following command:
It was successful. Both vcf and gvcf were generated.
GPU version I run the following command:
It seems there are some errors and GPU was not used. These are the output (part of the output were removed due to the limit of the characters of this post):
My system is Ubuntu 22.04. I have two GPUs.
nvidia-smi
I did run
export CUDA_VISIBLE_DEVICES=0
before running deepvariant, as you suggested in this issue https://github.com/google/deepvariant/issues/761.Why the GPU was not used in my run? Any help would be greatly appreciated!