Open yiliu30 opened 11 months ago
We don't have a concrete roadmap for that but any suggestions are welcomed. Do you have any particular features in mind you would like to see in OneFlow regarding Kubernetes?
The usage of oneflow with k8s should completely be the same as pytorch with k8s.
you can use oneflow with pytorch-operator https://github.com/kubeflow/pytorch-operator
We don't have a concrete roadmap for that but any suggestions are welcomed. Do you have any particular features in mind you would like to see in OneFlow regarding Kubernetes?
Thank you for your reply. Is it possible to port the supported distributed strategies to Kubernetes? For example, can we perform model parallelism across multiple pods?
The usage of oneflow with k8s should completely be the same as pytorch with k8s.
you can use oneflow with pytorch-operator https://github.com/kubeflow/pytorch-operator
Thanks for your guidance, this repo has been archived and the functionalities have been merged into the training-operator. It seems that these APIs are designed specifically for frameworks such as PyTorch and PaddlePaddle...
Description
I am curious to know if there are any ongoing or future efforts to make OneFlow compatible with Kubernetes for better orchestration and management of machine learning workloads. If such integration is already underway, could you provide some details on the progress? thx
Alternatives
No response