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

Fix missing ds_id bug #5824

Merged
merged 5 commits into from
Aug 14, 2024
Merged

Fix missing ds_id bug #5824

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
c57747e
Fix missing ds_id bug
tjruwase Aug 3, 2024
c07cae8
Merge branch 'master' into olruwase/ds_5495
loadams Aug 5, 2024
030e37a
Merge branch 'master' into olruwase/ds_5495
loadams Aug 6, 2024
8bd4509
Merge branch 'master' into olruwase/ds_5495
tjruwase Aug 9, 2024
8d75bd8
Merge branch 'master' into olruwase/ds_5495
tjruwase Aug 14, 2024
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
8 changes: 5 additions & 3 deletions deepspeed/runtime/zero/stage3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -821,10 +821,14 @@ def _create_fp32_partitions(self):

for i, tensor in enumerate(self.fp16_partitioned_groups_flat):
num_elements = self.fp16_partitioned_groups_flat_numel[i]
ds_id_begin = str(self.fp16_partitioned_groups_flat_id[i][0])
ds_id_end = str(self.fp16_partitioned_groups_flat_id[i][-1])
ds_id = ds_id_begin + '_' + ds_id_end

# a partition of the fp32 master weights that will be updated by this process
if self._swappable_optimizer_subgroup(i):
self.fp32_partitioned_groups_flat.append(torch.Tensor())
self.fp32_partitioned_groups_flat[i].ds_id = ds_id
nvme_memory_usage += (fp32_element_size * num_elements)
num_swappable_partitions += 1

Expand Down Expand Up @@ -861,11 +865,9 @@ def _create_fp32_partitions(self):
else:
self.fp32_partitioned_groups_flat.append(self.fp16_partitioned_groups_flat[i].to(
self.device).clone().float().detach())
self.fp32_partitioned_groups_flat[i].ds_id = ds_id

self.fp32_partitioned_groups_flat[i].requires_grad = True # keep this in case internal optimizer uses it
ds_id_begin = str(self.fp16_partitioned_groups_flat_id[i][0])
ds_id_end = str(self.fp16_partitioned_groups_flat_id[i][-1])
self.fp32_partitioned_groups_flat[i].ds_id = ds_id_begin + '_' + ds_id_end

if len(swappable_fp32_tensors) > 0:
self.optimizer_swapper.initialize_parameters(parameters=swappable_fp32_tensors,
Expand Down
Loading