Error: Manifest for floydhub/dl-docker:gpu not found
petecog opened this issue · comments
Hi - looks like the perfect docker build!
I followed your steps, but after the build, when I came to run it:
$ nvidia-docker run -it -p 8888:8888 -p 6006:6006 -v /media/data1:/root/sharedfolder floydhub/dl-docker:gpu bash
I got
nvidia-docker | 2017/04/12 11:35:25 Error: Error response from daemon: manifest for floydhub/dl-docker:gpu not found
I'm not sure if this is a docker / nvidia-docker / your build issue. But any steer you might have very much welcomed!
I notice that floydhub/dl-docker is not appearing as a repository when I run nvidia-docker images, but there is 'something' there, following the build. Rep and Tag = <none>
REPOSITORY TAG IMAGE ID CREATED SIZE
<none> <none> 9970b532e83a 55 minutes ago 10.5 GB nvidia/cuda 8.0-cudnn5-devel-ubuntu14.04 2b185ca91cd9 11 days ago 1.79 GB
Could the build have failed to 'register' the repository in some way?
Cancel my last! on a re-build it seems to work. Great set up. Many thanks.
One suggestion - PyMC3 would be a great addition, but there is already a lot going on in there!
Hi, @petecog, how do you solve this? I encounter the same error with you? What do you mean by "re-built"?
Thanks!
@tarvos21 It won't just resolve by rerun the docker build command. There is a missing flag for cmake in Dockerfile you need to add. Please refer to https://github.com/floydhub/dl-docker/issues/50
@hayday100 The issue seems closed and wrong link. Could you specify what flag to put and where, please?
@petecog Could you help with that, please?
hi @saiprashanths - you message got through.
Happy to close the issue. I've not been using the build since about May 17, so my memory is sketchy. I think I just removed the docker image and rebuilt from scratch and it worked.
If I recall, I stopped trying to use dockerised python because I couldn't work out a way of getting it to write to shared volumes as anything other than root.
@petecog I think it was me(about the message).
Could you maybe recommend a way/image you prefer to use when working with Deep Learning on Docker if any?