There are some images in https://hub.docker.com/u/pytorch. Is PyTorch installed in such image? If so, how to use it. If not, what is the purpose of such images.
[It is https://hub.docker.com/u/pytorch official docker instead of official nvidia-docker.]
BTW: what is pytorch/dockerfile designed for?
That confuses me a lot.
Yes, PyTorch is installed in these containers.
No, they are not maintained by NVIDIA.
You can find more information on docker containers here.
The Dockerfile is used to build the container.
Thanks for your reply. It seems some are installed pytorch while the other are not.
For example, https://hub.docker.com/r/pytorch/manylinux-cuda102 is just installed cuda toolkits without pytorch.
Sorry for not being clear. Some of these containers are used as the base e.g. to build the binaries etc.
The framework is installed in pytorch/pytorch using the latest stable version (1.6.0).
Much thanks, I understand it.
Summarize,
1.something like https://hub.docker.com/r/pytorch/manylinux-cuda102 is base docker and some kits in it.
2.something like https://hub.docker.com/r/pytorch/pytorch is docker with pytorch installed without source code.
3.https://ngc.nvidia.com/catalog/containers/nvidia:pytorch is maintained by nv and pytorch/cuda-kits installed. And pytorch code in it.