Is it possible to access gpu by using an nvidia docker image instead of the base python image? Will the pipeline able to access the gpu given we are running a container in a container?
I have been able to use bitbucket pipelines runners successfully for close to a year now with much success! Excellent work on this.
I am starting to find issues with some larger pipelines now, though. I'm finding that I have limited space in /tmp (which is by default used as the working directory), but the mkfifo command for the build_result file causes issues if I use my larger storage. Is there any way to configure the runner to use a larger storage for the working directory but a mkfifo-supported storage for just tmp (or ideally build_result)?
27 comments