How to set specified stage to specified device in pipeline parallelism? #5768
Unanswered
ShuaipengWu
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I want to set embedding layer (stage 0) to CPU to compute while other stages are still computed on GPU in deepspeed pipeline parallelism,so whether deepspeed supports this kind of mode or not？I also noticed that ZeRO Offload can conduct caculation with different kinds of devices,but I did not find how to set specified stage to CPU. Is there someone who try this?
Beta Was this translation helpful? Give feedback.
All reactions