epl单机单卡和单机多卡训练step如何理解

单机单卡：
启动命令：TF_CONFIG='{"cluster":{"worker":["127.0.0.1:49119"]},"task":{"type":"worker","index":0}}' CUDA_VISIBLE_DEVICES=0 bash ./scripts/train_dp.sh

单机双卡：
启动命令：TF_CONFIG='{"cluster":{"worker":["127.0.0.1:49119"]},"task":{"type":"worker","index":0}}' CUDA_VISIBLE_DEVICES=0,1 bash ./scripts/train_dp.sh

代码修改了一下：去掉了last_step限制，数据集repeat=10，将txt改为py，可执行。
[resnet_dp.txt](https://github.com/alibaba/EasyParallelLibrary/files/12671704/resnet_dp.txt)

想请教下，这个如何理解呢？每个卡分别跑了10step？

Start with scripts/train_dp.sh and the attached resnet_dp.txt/resnet_dp.py example, then compare the single-GPU and two-GPU commands with their reported steps. Document how dataset repeat, worker configuration, and per-device execution affect the step count; done means the issue has a clear explanation supported by the observed runs.

Written by the indexing model from the issue text.