No function of Docker container for DeepVariant

Question

[Tobi]

Hi,

I tried to install the docker container for DeepVariant as it is described in the documentation. But when I execute the commands in the container nothing happens. It does not even print an error message. dmesg gives me the following line after the execution of ./opt/deepvariant/bin/make_examples: traps: python[44779] trap invalid opcode ip:7f87f0fe6891 sp:7ffcc9fbf3d0 error:0 in libtensorflow_framework.so[7f87f0bff000+c00000] Sadly, I could not find any helpful information on the internet regarding this output.

Could anyone help me to make the software run?

Thanks,

Tobi

[GenoMax]

Tagging: mdepristo

[arostami]

Hi Tobi,

Could you please let us know what OS and environment you are using for running the docker image?

Thanks, Asha

[GenoMax]

Please use ADD COMMENT/ADD REPLY when responding to existing posts to keep threads logically organized.

[Tobi]

Hi Asha,

I'm using an Ubuntu 16.04 to run the docker image. The computer has 24 cores and 48 GB of RAM so this should be enough I guess. I am not using a GPU.

Does this help you or do you need more information?

Best, Tobi

[arostami]

Thanks for providing the info, Tobi. This is very strange and it should work with Ubuntu 16. I haven't seen this error before, but it looks like there may be a hardware configuration issue.

• Could you please confirm that you can run the regular binaries (outside of docker) following the quick start instructions? https://github.com/google/deepvariant/blob/r0.4/docs/deepvariant-quick-start.md
• Could you please retry with IMAGE_VERSION=0.4.1 (just released)?
• Lastly, can you try running a simple python program (shown below) inside the docker image? I want to ensure that tensorflow can properly run in your environment.

Sample test code:

import tensorflow as tf
def main(argv=()):
  print 'ok'
if __name__ == '__main__':
  tf.app.run()

[Tobi]

I tested a few things but sadly I got no positive results:

• the regular binaries do not run correctly outside of docker. When I run them nothing happens and using dmesg I get a similar error to the one I get inside of docker.
• Using the new image version changes nothing. Neither when using docker nor when not using it.
• While your test code runs on the machine without docker, inside of the docker container the import line gives me Illegal instruction (core dumped)...

Now I do not have any Idea what I could change to make DeepVariant run. I will also have a look at this issue, I think it is related to my case.

[Tobi]

Sadly, the linked issue did not help me. Therefore I need further help. For now I will try it on another machine.

Answer 1

[alaincoletta]

Hi Tobi, we implemented the DeepVariant pipeline with Docker and Nextflow here https://lifebit.page.link/gnbR . You can try it with example parameters.

Would love your feedback on this.

Thanks!