transformers
b56e1d68
- Merge branch 'fsdp2-ep-2d-mesh' into ep-token-dispatch
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
18 days ago
Merge branch 'fsdp2-ep-2d-mesh' into ep-token-dispatch
References
#48204 - [Umbrella] FSDP2 + expert-parallel token dispatch, with sharded-loading prefetch
#48518 - Expert-parallel token dispatch: each rank trains on its own part of the batch
#48792 - Add Expert parallel
Author
qgallouedec
Parents
db788602
a6f5824a
Loading