intel / intel/intel-optimization-for-horovod
Docker use
Open
enhancement
- Dominant language
- Python
- Stars
- 5
- Forks
- 5
- PR merge metrics
- No merged PRs in 30d
Description
Container "gpu-max" from here https://hub.docker.com/r/intel/intel-extension-for-tensorflow/ supports horovod execution with GPUs (you need to set a few env variables https://github.com/intel/intel-extension-for-tensorflow/issues/29). With it xpu examples run fine as long as you expose GPUs and enable host IPC (docker's --ipc=host). Exposing GPUs is obvious, but I feel the IPC is something not all people know of.
Maybe in the README, there could be a note about requirement(s) for the execution environment, such as the IPC.
Contributor guide
Assessment
This issue has not been assessed yet.