Le-yi

icon company Mahō Shōjo特別作戦班です icon location Beijing icon location Scale-up during the day, scale-out at night.

Results 2 comments of Le-yi

[BUG] Cannot use --hostfile to start multi-node training in Docker.

Thank you very much for your guidance @GuanhuaWang . But I face a new issue. When training on two machines, each with 8 A100 GPUs, I encountered the following error....

[BUG] Cannot use --hostfile to start multi-node training in Docker.

> [@Ind1x1](https://github.com/Ind1x1)- 你能分享你的共享家庭内存大小吗？有时当内存设置得太低时，就会发生这种情况。 Sorry it took so long to reply, I configured docker with a shared memory size of 100GB on each machine @loadams