Skip to content

Offload all gradients to nvme #2282

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 5 commits into from
Sep 5, 2022
Merged

Offload all gradients to nvme #2282

Changes from 3 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
843821f
Offload all gradients to nvme
tjruwase Sep 1, 2022
de46473
Merge branch 'master' into olruwase/nvme_finetune
tjruwase Sep 1, 2022
aed6eaf
Merge branch 'master' into olruwase/nvme_finetune
tjruwase Sep 1, 2022
1b2ab39
Merge branch 'master' into olruwase/nvme_finetune
tjruwase Sep 4, 2022
51f817f
Merge branch 'master' into olruwase/nvme_finetune
tjruwase Sep 4, 2022
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
4 changes: 2 additions & 2 deletions deepspeed/runtime/zero/stage3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1239,6 +1239,8 @@ def complete_grad_norm_calculation_for_cpu_offload(self, params):
def __partition_grads(self,
params_to_release: List[Parameter],
grad_partitions: List[Tensor]) -> None:
offload_fp32_gradients = {}
offload_fp32_offsets = {}
for param, grad_partition in zip(params_to_release, grad_partitions):
if param.partition_numel() * dist.get_rank(
self.dp_process_group) > param.ds_numel:
Expand Down Expand Up @@ -1280,8 +1282,6 @@ def __partition_grads(self,
# offload the gradient partition if applicable
if self.offload_optimizer:
i, dest_offset, _ = self.grad_position[self.get_param_id(param)]
offload_fp32_gradients = {}
offload_fp32_offsets = {}

if self.is_gradient_accumulation_boundary:
self.norm_for_param_grads[self.get_param_id(
Expand Down