transformers
[`distributed`] Add context parallelism: `DistributedConfig(cp_size=N)`
#50496
Open

[`distributed`] Add context parallelism: `DistributedConfig(cp_size=N)` #50496

3outeille wants to merge 8 commits into cp-1-implementations from cp-3-cp-size
3outeille
3outeille [distributed] Add context parallelism sizes to DistributedConfig
faf2d3e7
3outeille [distributed] Add the cp axis to the device mesh and shard FSDP2 over…
931c06b8
3outeille [distributed] Refuse models whose layers or head context parallelism …
cb227c6d
3outeille [distributed] Split the sequences across the cp group when the model …
8051af49
3outeille [distributed] Refuse context parallelism in Trainer
95e96cb3
3outeille [distributed] Document context parallelism in the load and config doc…
5964f5d8
3outeille [distributed] Test context parallelism against the same weights on fu…
635bb98e
HuggingFaceDocBuilderDev
3outeille [distributed] Refuse context parallelism for attention outside the at…
6b2c9674

Login to write a write a comment.

Login via GitHub

Reviewers
No reviews
Assignees
No one assigned
Labels
Milestone