Update theano about dockerfiles HOT 4 CLOSED

Also, it would be great if you can update the device flag to cuda instead of gpu and lib.cnmem to gpuarray.preallocate. These correspond to the new backend. The old backend has now been deleted.
https://hub.docker.com/r/floydhub/theano/~/dockerfile/

from dockerfiles.

Thank you very much for updating the version and adding support to libgpuarray. I use the bleeding edge version of floyd-cli but I am not able to use this version of theano when i use theano:py2 environment. Could you help me out with that ? @houqp @narenst
Thanks!

from dockerfiles.

Hi @Sentient07 , I have already added pygpu through manual compilation: https://github.com/floydhub/dockerfiles/blob/master/dl/theano/0.9.0/Dockerfile-py3.gpu#L31.

I am still testing the new image to make sure it's working properly. If everything goes well, floydhub will make it available to the CLI in two weeks I think.

from dockerfiles.

Hi @Sentient07 , we have released a new set of chainer images. You can test it out by upgrading your floyd cli and run the experiment with --env chainer-2.0 or --env chainer-1.23 arguments.

from dockerfiles.