Difference between huggingface deepspeed integration and accelerate deepspeed-plugin

[jerryli1981]

huggingface transformer support deepspeed (https://huggingface.co/docs/transformers/main_classes/deepspeed )

accelerate also support deepspeed via ds-plugin, so what is design thoughts besides no trainer flexibility?

[jerryli1981]

Another quick question is "BatchSamplerShard" in data_loader.py is functionally equal to torch.utils.data.distributed.DistributedSampler ?

[sgugger]

Please use the forums to ask questions as we keep the issues for bugs and feature requests only. The Trainer supports deepspeed but Accelerate is designed for people who don't want to use a Trainer. It also supports deepseed for people who want to use that library and retain full control over their training loop.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.