Skip to content

[zero partitioning] fix double linear override; spelling #954

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

Merged
tjruwase merged 1 commit into from
Apr 14, 2021
Merged

[zero partitioning] fix double linear override; spelling #954

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
9 changes: 1 addition & 8 deletions deepspeed/runtime/zero/partition_parameters.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -191,7 +191,7 @@ def _init_subclass(cls, **kwargs):

if self.mem_efficient_linear:
print_rank_0(
f"Your linear layers are being patched with more memory efficient version. This will persit unless manually reset.",
"nn.functional.linear has been overridden with a more memory efficient version. This will persist unless manually reset.",
force=True)
self.linear_bk = torch.nn.functional.linear
torch.nn.functional.linear = LinearFunctionForZeroStage3.apply
Expand Down Expand Up @@ -361,13 +361,6 @@ def get_model():
self._convert_to_deepspeed_param(param)
param.partition()

if mem_efficient_linear:
print_rank_0(
f"Your linear layers are being patched with more memory efficient version. This will persit unless manually turned reset.",
force=True)
self.linear_bk = torch.nn.functional.linear
torch.nn.functional.linear = LinearFunctionForZeroStage3.apply

def _post_init_method(self, module):
#see_memory_usage(f"Before converting parmas in {module.__class__.__name__}", force=False)
print_rank_0(f'Converting Params in {module.__class__.__name__}', force=False)
Expand Down