Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

share inflight registry between PartitionedParameterCoordinators #3462

Merged
merged 7 commits into from
May 15, 2023
Merged

share inflight registry between PartitionedParameterCoordinators #3462

Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
993c304
share inflight registry between PartitionedParameterCoordinator
HeyangQin May 5, 2023
30c91c0
bound registry to model
HeyangQin May 5, 2023
16a2122
make InflightParamRegistry standalone
HeyangQin May 5, 2023
8413e7f
Merge branch 'master' into HeyangQin/shared_inflightregistry
HeyangQin May 9, 2023
53f11ef
fix format
HeyangQin May 12, 2023
951cc25
Merge branch 'master' into HeyangQin/shared_inflightregistry
HeyangQin May 12, 2023
0cd0887
Merge branch 'master' into HeyangQin/shared_inflightregistry
tjruwase May 15, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
bound registry to model
  • Loading branch information
@HeyangQin
HeyangQin committed May 5, 2023
commit 30c91c04805292ce3702915e46eb4caa95548931
5 changes: 4 additions & 1 deletion deepspeed/runtime/zero/parameter_offload.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -243,7 +243,10 @@ def __init__(self,
self._max_reuse_distance_in_numel = int(max_reuse_distance)
self._max_available_parameters_in_numel = int(max_live_parameters)
self.__allgather_stream = get_accelerator().Stream() if overlap_comm else get_accelerator().default_stream()
self.__inflight_param_registry = PartitionedParameterCoordinator.__InflightParamRegistry()

if not hasattr(module, "ds_inflight_param_registry"):
module.ds_inflight_param_registry = PartitionedParameterCoordinator.__InflightParamRegistry()
self.__inflight_param_registry = module.ds_inflight_param_registry

self.forward_hooks = []
self.backward_hooks = []
Expand Down