LiSu

I cannot repeat this problem, following this [tutorial](https://graph-learn.readthedocs.io/en/latest/en/dgs/tutorial.html) I can successfully deploy DGS. Maybe you can check if helm/k8s works well in your environment.

We are working on supporting multi-server training using GLTorch in GraphScope now, after that it should be OK to deprecate GL from GraphScope. @yecol

> I'm not sure I follow. Is there a specific issue you're facing? Thanks for the quick reply! The background is that we are trying to build a RPC framework...

Hi Luca @lw, I think we found the answer for the above question from your comment in [issue 405 ](https://github.com/pytorch/tensorpipe/issues/405): > These two sides are used as ringbuffers. When a...

GraphScope基于分布式GNN训练框架graphlearn-for-pytorch (GLTorch)支持大规模分布式GNN训练。GLTorch在模型层和PyG兼容，支持将PyG GNN训练扩展到分布式大图。为了支持在大于机器可用内存大小的图上训练GNN，PyG引入了一套可插拔的Remote Backend机制，即通过FeatureStore 和 GraphStore等抽象，支持第三方图存储引擎和PyG的对接。其中FeatureStore允许用户利用存储在远程的节点/边特征，GraphStore允许用户利用存储在远程的图结构信息，两者结合支持基于远端存储的GNN训练扩展。本项目旨在通过实现基于Graphscope的PyG Remote Backend，更进一步简化GraphScope和PyG的对接方式，提供对PyG用户友好的基于GraphScope进行分布式GNN训练的产品使用体验。 产出： - 基于GLTorch当前架构，设计FeatureStore 和 GraphStore在GraphScope上的实现方案 - 完成Remote Backend整体实现，在GraphScope上提供基于PyG Remote Backend的分布式训练示例 难度： 初级 技术要求：熟练使用Python语言，熟悉C++

> @drcanchi @LiSu can you please review these changes? LGTM, thanks! @hiwotadese @nv-rborkar could you please check this PR?

> recheck Thanks for the PR! @hiwotadese @nv-rborkar could you please check this PR?